Muse Ember
Explore Muse Ember from Meta: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.
Compare Muse Ember with other models →Compare subscription plansExplore data coverage
Pricing · OpenRouter
Dated cached OpenRouter rates in USD per 1M tokens. Open the dashboard for live enhancements. Per-metric endpoint minima can refer to different providers; they are not a guaranteed combined rate from one endpoint.
| Tier | Input / 1M | Output / 1M | Cached input / 1M | Cache write / 1M | Date & source |
|---|---|---|---|---|---|
| Default | $0.1 | $0.5 | $0.004 | $0.125 | 2026-11-10 · OpenRouter source |
Recorded pricing notes
OpenRouter catalog list price read 2026-11-10 (route created 2026-11-09); matches Meta's list price for prompts up to 200K tokens. No contributor tier. The snapshot workflow replaces this with healthy-endpoint minima; min-healthy endpoint minima 2026-11-10
Official / vendor benchmarks
Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.
| Benchmark / evaluator | Headline score | Evidence |
|---|---|---|
| AutomationBench-AA (vendor-reported) | 37.8%max effort | The recorded result & source37.8% · raw 37.8 % Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. AutomationBench-AA, Artificial Analysis, strict pass@1. https://research.meta.ai/blog/muse-ember |
| Terminal-Bench 4.0 | 40.9%max effort | The recorded result & source40.9% · raw 40.9 % Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. Harness: mini-swe-agent (Meta: each model in its best published harness). https://research.meta.ai/blog/muse-ember |
| GDP.PDF (vendor reports) | 22.8%max effort | The recorded result & source22.8% · raw 22.8 % Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. GDP.pdf-AA, Artificial Analysis. https://research.meta.ai/blog/muse-ember |
| OSWorld V2.1 (108 tasks, partial reward) | 69.5%max effort | The recorded result & source69.5% · raw 69.5 % Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. OSWorld 2.1, Meta internal evaluation, partial-credit reward. https://research.meta.ai/blog/muse-ember |
| OSWorld V2.1 (108 tasks, binary) | 33.8%max effort | The recorded result & source33.8% · raw 33.8 % Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. OSWorld 2.1, Meta internal evaluation, binary completion. https://research.meta.ai/blog/muse-ember |
| AA Intelligence Index v5.0 (vendor-cited) | 36max effort | The recorded result & source36 · raw 36 index Headline · max effort · Own vendor Source/record date: 2026-11-09 Meta Muse Ember launch post (9 November 2026), max reasoning. Artificial Analysis Intelligence Index v5 as published 9 November 2026; Meta notes cost per task about $0.14 at list pricing. https://research.meta.ai/blog/muse-ember |
Independent evaluators
Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.
| Benchmark / evaluator | Headline score | Evidence |
|---|---|---|
| Excel Modeling Benchmark v1.1 · Vals AI | 72.169%max effort | The recorded result & source72.169% · raw 72.169 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Cost per task · Vals AI | $0.091434/taskmax effort | The recorded result & source$0.091434/task · raw 0.091434 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Task latency · Vals AI | 289.841 smax effort | The recorded result & source289.841 s · raw 289.841 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Template · Vals AI | 76.208%max effort | The recorded result & source76.208% · raw 76.208 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Scratch · Vals AI | 67.980%max effort | The recorded result & source67.980% · raw 67.98 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · LBO · Vals AI | 60.722%max effort | The recorded result & source60.722% · raw 60.722 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · DCF · Vals AI | 62.577%max effort | The recorded result & source62.577% · raw 62.577 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · M&A · Vals AI | 80.440%max effort | The recorded result & source80.440% · raw 80.44 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Operating · Vals AI | 71.212%max effort | The recorded result & source71.212% · raw 71.212 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Comps · Vals AI | 75.732%max effort | The recorded result & source75.732% · raw 75.732 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Dataroom · Vals AI | 87.359%max effort | The recorded result & source87.359% · raw 87.359 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Excel Modeling Benchmark v1.1 · Misc · Vals AI | 67.140%max effort | The recorded result & source67.140% · raw 67.14 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/emb-v1-1 |
| Vals Finance Index v1 · Vals AI | 60.914%max effort | The recorded result & source60.914% · raw 60.914 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/finance_index |
| Vals Finance Index v1 · Cost per task · Vals AI | $0.059873/taskmax effort | The recorded result & source$0.059873/task · raw 0.059873 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/finance_index |
| Vals Finance Index v1 · Task latency · Vals AI | 156.441 smax effort | The recorded result & source156.441 s · raw 156.441 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/finance_index |
| Vals Finance Index v1 · Finance Agent v2.1 · Vals AI | 63.834%max effort | The recorded result & source63.834% · raw 63.834 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/finance_index |
| Vals Finance Index v1 · EMB v1.1 · Vals AI | 72.169%max effort | The recorded result & source72.169% · raw 72.169 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/finance_index |
| Vals Finance Index v1 · Tax Agent all-pass accuracy · Vals AI | 38.233%max effort | The recorded result & source38.233% · raw 38.233 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/finance_index |
| Finance Agent v2.1 · Vals AI | 63.834%max effort | The recorded result & source63.834% · raw 63.834 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Cost per task · Vals AI | $0.056396/taskmax effort | The recorded result & source$0.056396/task · raw 0.056396 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Task latency · Vals AI | 89.351 smax effort | The recorded result & source89.351 s · raw 89.351 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · All-Pass Accuracy · Vals AI | 50.852%max effort | The recorded result & source50.852% · raw 50.852 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · General Qualitative Analysis · Vals AI | 77.126%max effort | The recorded result & source77.126% · raw 77.126 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · General Quantitative Analysis · Vals AI | 76.431%max effort | The recorded result & source76.431% · raw 76.431 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Market Analysis · Vals AI | 64.933%max effort | The recorded result & source64.933% · raw 64.933 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Comparables · Vals AI | 60.249%max effort | The recorded result & source60.249% · raw 60.249 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Precedents · Vals AI | 32.975%max effort | The recorded result & source32.975% · raw 32.975 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Adjustments · Vals AI | 64.279%max effort | The recorded result & source64.279% · raw 64.279 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Earnings Analysis · Vals AI | 87.158%max effort | The recorded result & source87.158% · raw 87.158 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Disclosure Analysis · Vals AI | 75.066%max effort | The recorded result & source75.066% · raw 75.066 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| Finance Agent v2.1 · Financial Modeling · Vals AI | 36.286%max effort | The recorded result & source36.286% · raw 36.286 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/fabv2-1 |
| HLAB v1.1 (Vals) · Task Pass Rate · Vals AI | 25.000%max effort | The recorded result & source25.000% · raw 25 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Cost per task · Vals AI | $0.118448/taskmax effort | The recorded result & source$0.118448/task · raw 0.118448 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task latency · Vals AI | 353.605 smax effort | The recorded result & source353.605 s · raw 353.605 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Antitrust Competition · Vals AI | 10.000%max effort | The recorded result & source10.000% · raw 10 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Arbitration International Dispute Resolution · Vals AI | 20.000%max effort | The recorded result & source20.000% · raw 20 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Banking Finance · Vals AI | 20.000%max effort | The recorded result & source20.000% · raw 20 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Bankruptcy Restructuring · Vals AI | 30.000%max effort | The recorded result & source30.000% · raw 30 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Capital Markets · Vals AI | 40.000%max effort | The recorded result & source40.000% · raw 40 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Corporate Governance · Vals AI | 50.000%max effort | The recorded result & source50.000% · raw 50 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Corporate Ma · Vals AI | 30.000%max effort | The recorded result & source30.000% · raw 30 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Data Privacy Cybersecurity · Vals AI | 40.000%max effort | The recorded result & source40.000% · raw 40 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Emerging Companies Venture Capital · Vals AI | 30.000%max effort | The recorded result & source30.000% · raw 30 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Employment Labor · Vals AI | 10.000%max effort | The recorded result & source10.000% · raw 10 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Energy Natural Resources · Vals AI | 60.000%max effort | The recorded result & source60.000% · raw 60 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Environmental Esg · Vals AI | 20.000%max effort | The recorded result & source20.000% · raw 20 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Funds Asset Management · Vals AI | 50.000%max effort | The recorded result & source50.000% · raw 50 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Healthcare Life Sciences · Vals AI | 40.000%max effort | The recorded result & source40.000% · raw 40 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Immigration · Vals AI | 30.000%max effort | The recorded result & source30.000% · raw 30 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Insurance · Vals AI | 0.000%max effort | The recorded result & source0.000% · raw 0 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Intellectual Property · Vals AI | 20.000%max effort | The recorded result & source20.000% · raw 20 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · International Trade Sanctions · Vals AI | 30.000%max effort | The recorded result & source30.000% · raw 30 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Litigation Dispute Resolution · Vals AI | 10.000%max effort | The recorded result & source10.000% · raw 10 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Real Estate · Vals AI | 0.000%max effort | The recorded result & source0.000% · raw 0 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Structured Finance Securitization · Vals AI | 20.000%max effort | The recorded result & source20.000% · raw 20 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Tax · Vals AI | 0.000%max effort | The recorded result & source0.000% · raw 0 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · Trusts Estates Private Client · Vals AI | 40.000%max effort | The recorded result & source40.000% · raw 40 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Task Pass Rate · White Collar Defense Investigations · Vals AI | 0.000%max effort | The recorded result & source0.000% · raw 0 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Vals AI | 95.546%max effort | The recorded result & source95.546% · raw 95.546 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Antitrust Competition · Vals AI | 96.196%max effort | The recorded result & source96.196% · raw 96.196 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Arbitration International Dispute Resolution · Vals AI | 96.057%max effort | The recorded result & source96.057% · raw 96.057 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Banking Finance · Vals AI | 96.455%max effort | The recorded result & source96.455% · raw 96.455 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Bankruptcy Restructuring · Vals AI | 95.502%max effort | The recorded result & source95.502% · raw 95.502 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Capital Markets · Vals AI | 95.882%max effort | The recorded result & source95.882% · raw 95.882 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Corporate Governance · Vals AI | 95.890%max effort | The recorded result & source95.890% · raw 95.89 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Corporate Ma · Vals AI | 97.267%max effort | The recorded result & source97.267% · raw 97.267 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Data Privacy Cybersecurity · Vals AI | 98.410%max effort | The recorded result & source98.410% · raw 98.41 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Emerging Companies Venture Capital · Vals AI | 95.734%max effort | The recorded result & source95.734% · raw 95.734 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Employment Labor · Vals AI | 91.573%max effort | The recorded result & source91.573% · raw 91.573 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Energy Natural Resources · Vals AI | 98.680%max effort | The recorded result & source98.680% · raw 98.68 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Environmental Esg · Vals AI | 93.907%max effort | The recorded result & source93.907% · raw 93.907 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Funds Asset Management · Vals AI | 93.732%max effort | The recorded result & source93.732% · raw 93.732 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Healthcare Life Sciences · Vals AI | 97.315%max effort | The recorded result & source97.315% · raw 97.315 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Immigration · Vals AI | 94.727%max effort | The recorded result & source94.727% · raw 94.727 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Insurance · Vals AI | 94.118%max effort | The recorded result & source94.118% · raw 94.118 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Intellectual Property · Vals AI | 95.196%max effort | The recorded result & source95.196% · raw 95.196 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · International Trade Sanctions · Vals AI | 97.012%max effort | The recorded result & source97.012% · raw 97.012 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Litigation Dispute Resolution · Vals AI | 95.301%max effort | The recorded result & source95.301% · raw 95.301 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Real Estate · Vals AI | 95.918%max effort | The recorded result & source95.918% · raw 95.918 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Structured Finance Securitization · Vals AI | 94.839%max effort | The recorded result & source94.839% · raw 94.839 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Tax · Vals AI | 92.642%max effort | The recorded result & source92.642% · raw 92.642 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · Trusts Estates Private Client · Vals AI | 95.229%max effort | The recorded result & source95.229% · raw 95.229 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| HLAB v1.1 (Vals) · Criteria Pass Rate · White Collar Defense Investigations · Vals AI | 95.455%max effort | The recorded result & source95.455% · raw 95.455 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/hlab |
| Vals Index v2.2 · Vals AI | 60.475%max effort | The recorded result & source60.475% · raw 60.475 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Cost per task · Vals AI | $0.15305/taskmax effort | The recorded result & source$0.15305/task · raw 0.15305 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Task latency · Vals AI | 902.371 smax effort | The recorded result & source902.371 s · raw 902.371 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Vibe Code Bench v1.1 · Vals AI | 83.703%max effort | The recorded result & source83.703% · raw 83.703 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: OpenHands. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Finance Agent v2.1 · Vals AI | 63.834%max effort | The recorded result & source63.834% · raw 63.834 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · EMB v1.1 · Vals AI | 72.169%max effort | The recorded result & source72.169% · raw 72.169 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Code Migration v1.1 (index subset 50 CLI + 10 COBOL) · Vals AI | 47.108%max effort | The recorded result & source47.108% · raw 47.108 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Index Code Migration is the 60-task subset (50 CLI + 10 COBOL), not standalone Migration. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Legal Research · Vals AI | 49.038%max effort | The recorded result & source49.038% · raw 49.038 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · HLAB v1.1 · Vals AI | 25.000%max effort | The recorded result & source25.000% · raw 25 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Terminal-Bench 4.0 · Vals AI | 39.394%max effort | The recorded result & source39.394% · raw 39.394 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Index v2.2 · Tax Agent · Vals AI | 38.233%max effort | The recorded result & source38.233% · raw 38.233 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vals_index |
| Vals Legal Index v1 · Vals AI | 75.686%max effort | The recorded result & source75.686% · raw 75.686 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Cost per task · Vals AI | $0.068939/taskmax effort | The recorded result & source$0.068939/task · raw 0.068939 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Task latency · Vals AI | 217.157 smax effort | The recorded result & source217.157 s · raw 217.157 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Business & Commercial · Vals AI | 73.346%max effort | The recorded result & source73.346% · raw 73.346 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Civil Negligence · Vals AI | 71.914%max effort | The recorded result & source71.914% · raw 71.914 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Other Civil · Vals AI | 75.060%max effort | The recorded result & source75.060% · raw 75.06 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Real Estate · Vals AI | 95.918%max effort | The recorded result & source95.918% · raw 95.918 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Labor & Employment · Vals AI | 91.573%max effort | The recorded result & source91.573% · raw 91.573 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Wills, Estates & Trusts · Vals AI | 95.229%max effort | The recorded result & source95.229% · raw 95.229 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Family · Vals AI | 27.273%max effort | The recorded result & source27.273% · raw 27.273 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Criminal · Vals AI | 73.166%max effort | The recorded result & source73.166% · raw 73.166 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Tax Agent · Vals AI | 71.353%max effort | The recorded result & source71.353% · raw 71.353 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · Legal Research · Vals AI | 49.038%max effort | The recorded result & source49.038% · raw 49.038 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Vals Legal Index v1 · HLAB Criteria Pass · Vals AI | 95.546%max effort | The recorded result & source95.546% · raw 95.546 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_index |
| Legal Research Bench · Vals AI | 49.038%max effort | The recorded result & source49.038% · raw 49.038 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Cost per task · Vals AI | $0.056581/taskmax effort | The recorded result & source$0.056581/task · raw 0.056581 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Task latency · Vals AI | 207.736 smax effort | The recorded result & source207.736 s · raw 207.736 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Administrative / Regulatory · Vals AI | 54.545%max effort | The recorded result & source54.545% · raw 54.545 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Business & Commercial · Vals AI | 51.163%max effort | The recorded result & source51.163% · raw 51.163 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Civil Litigation · Vals AI | 48.148%max effort | The recorded result & source48.148% · raw 48.148 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Constitutional / Civil Rights · Vals AI | 42.105%max effort | The recorded result & source42.105% · raw 42.105 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Criminal · Vals AI | 50.877%max effort | The recorded result & source50.877% · raw 50.877 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Family · Vals AI | 27.273%max effort | The recorded result & source27.273% · raw 27.273 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Health · Vals AI | 70.000%max effort | The recorded result & source70.000% · raw 70 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · All-Pass Accuracy · Immigration · Vals AI | 45.455%max effort | The recorded result & source45.455% · raw 45.455 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Vals AI | 86.217%max effort | The recorded result & source86.217% · raw 86.217 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Administrative / Regulatory · Vals AI | 86.020%max effort | The recorded result & source86.020% · raw 86.02 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Business & Commercial · Vals AI | 83.679%max effort | The recorded result & source83.679% · raw 83.679 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Civil Litigation · Vals AI | 91.123%max effort | The recorded result & source91.123% · raw 91.123 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Constitutional / Civil Rights · Vals AI | 89.841%max effort | The recorded result & source89.841% · raw 89.841 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Criminal · Vals AI | 90.162%max effort | The recorded result & source90.162% · raw 90.162 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Family · Vals AI | 76.299%max effort | The recorded result & source76.299% · raw 76.299 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Health · Vals AI | 93.751%max effort | The recorded result & source93.751% · raw 93.751 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Legal Research Bench · Weighted Pass (citation-adjusted) · Immigration · Vals AI | 76.669%max effort | The recorded result & source76.669% · raw 76.669 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/legal_research |
| Tax Agent Bench Accuracy (all-pass, citation-adjusted) · Vals AI | 38.233%max effort | The recorded result & source38.233% · raw 38.233 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Cost per task · Vals AI | $0.031788/taskmax effort | The recorded result & source$0.031788/task · raw 0.031788 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Task latency · Vals AI | 90.131 smax effort | The recorded result & source90.131 s · raw 90.131 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Fact-Pattern Analysis · Vals AI | 44.186%max effort | The recorded result & source44.186% · raw 44.186 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Rule & Source Lookup · Vals AI | 53.333%max effort | The recorded result & source53.333% · raw 53.333 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Current & Temporal Analysis · Vals AI | 42.745%max effort | The recorded result & source42.745% · raw 42.745 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Numbers & Calculations · Vals AI | 43.222%max effort | The recorded result & source43.222% · raw 43.222 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Forms & Filings · Vals AI | 26.667%max effort | The recorded result & source26.667% · raw 26.667 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · All-Pass Accuracy · Controversy & Precedence Analysis · Vals AI | 16.667%max effort | The recorded result & source16.667% · raw 16.667 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench Weighted Pass (citation-adjusted) · Vals AI | 71.353%max effort | The recorded result & source71.353% · raw 71.353 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Fact-Pattern Analysis · Vals AI | 69.462%max effort | The recorded result & source69.462% · raw 69.462 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Rule & Source Lookup · Vals AI | 74.048%max effort | The recorded result & source74.048% · raw 74.048 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Current & Temporal Analysis · Vals AI | 73.183%max effort | The recorded result & source73.183% · raw 73.183 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Numbers & Calculations · Vals AI | 74.309%max effort | The recorded result & source74.309% · raw 74.309 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Forms & Filings · Vals AI | 81.190%max effort | The recorded result & source81.190% · raw 81.19 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Tax Agent Bench · Weighted Pass · Controversy & Precedence Analysis · Vals AI | 56.744%max effort | The recorded result & source56.744% · raw 56.744 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/tax_agent_bench |
| Terminal-Bench 4.0 · Vals AI | 39.394%max effort | The recorded result & source39.394% · raw 39.394 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Cost per task · Vals AI | $0.218381/taskmax effort | The recorded result & source$0.218381/task · raw 0.218381 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Task latency · Vals AI | 1581.448 smax effort | The recorded result & source1581.448 s · raw 1581.448 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Software · Vals AI | 46.296%max effort | The recorded result & source46.296% · raw 46.296 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Science · Vals AI | 35.714%max effort | The recorded result & source35.714% · raw 35.714 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · ML · Vals AI | 48.485%max effort | The recorded result & source48.485% · raw 48.485 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Operations · Vals AI | 33.333%max effort | The recorded result & source33.333% · raw 33.333 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Hardware · Vals AI | 13.333%max effort | The recorded result & source13.333% · raw 13.333 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Security · Vals AI | 33.333%max effort | The recorded result & source33.333% · raw 33.333 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Terminal-Bench 4.0 · Media · Vals AI | 50.000%max effort | The recorded result & source50.000% · raw 50 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: published Vals component protocol. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/terminal-bench-4 |
| Vibe Code Bench v1.1 · Vals AI | 83.703%max effort | The recorded result & source83.703% · raw 83.703 % Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: OpenHands. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. https://www.vals.ai/benchmarks/vibe-code |
| Vibe Code Bench v1.1 cost per task · Vals AI | $0.054457/taskmax effort | The recorded result & source$0.054457/task · raw 0.054457 USD/task Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: OpenHands. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/vibe-code |
| Vibe Code Bench v1.1 · Task latency · Vals AI | 167.856 smax effort | The recorded result & source167.856 s · raw 167.856 s Headline · max effort · Independent evaluator Source/record date: 2026-11-10 Harness: OpenHands. Exact Vals task/version and source model. Requested effort, provider and diagnostics are retained with this observation; evaluation date and checkpoint hash are unpublished. Source update date is not execution date. Quality standard error is not a confidence interval. A source revision does not by itself establish a model update. Workload diagnostic; quality uncertainty does not describe its uncertainty. Not a token tariff or controlled efficiency ranking. https://www.vals.ai/benchmarks/vibe-code |
Read how we select and source scores or the comparison guide.