Gemini 3.8 Flash Benchmarks, Specifications & Availability
Explore Gemini 3.8 Flash from Google: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.
Compare Gemini 3.8 Flash with other models →Explore data coverage
Published specifications
- Provider
- Access
- Proprietary
- License
- Proprietary
- Context window
- 1M
- Total parameters
- Not published
- Active parameters
- Not published
- Released
- 2026-09-03
- Modalities
- text, image, video, audio, file
- Family
- Gemini 3
Model card · Announcement · Website · OpenRouter
Model notes
Flash-tier workhorse for long-horizon SWE and agentic workflows.
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 | [object Object] | [object Object] | [object Object] | [object Object] | 2026-10-03 · OpenRouter source |
Recorded pricing notes
min-healthy endpoint minima 2026-10-01; discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); min-healthy endpoint minima 2026-10-01; discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); min-healthy endpoint minima 2026-10-02; discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); min-healthy endpoint minima 2026-10-03; discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied); discount flag 0.5 on Google AI Studio (undocumented, not applied)
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 |
|---|---|---|
| AA Coding Agent Index v1.4 | 61.2unknown effort | All 1 recorded result & sources61.2 · raw 61.2 index Headline · unknown effort · Peer vendor Source/record date: 2026-09-03 As reported in OpenAI GPT-6 Astra table https://openai.com/index/gpt-6-astra/ |
| BioMysteryBench | 88.8%unknown effort | All 1 recorded result & sources88.8% · raw 88.8 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 BioMysteryBench human-solvable https://deepmind.google/models/model-cards/gemini-3-8-flash |
| CharXiv Reasoning | 86.2%unknown effort | All 1 recorded result & sources86.2% · raw 86.2 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 CharXiv Reasoning no tools https://deepmind.google/models/model-cards/gemini-3-8-flash |
| DeepSWE v1.1 | 73.7%unknown effort | All 1 recorded result & sources73.7% · raw 73.7 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 DeepMind model card (prefer over OpenAI 73.8 if reconciling) https://deepmind.google/models/model-cards/gemini-3-8-flash |
| FrontierCode 1.1 Extended | 56.3%unknown effort | All 1 recorded result & sources56.3% · raw 56.3 % Headline · unknown effort · Peer vendor Source/record date: 2026-09-03 As reported in OpenAI GPT-6 Astra table (Gemini 3.8 Flash col) https://openai.com/index/gpt-6-astra/ |
| FrontierCode 1.1 Main | 43.6%unknown effort | All 1 recorded result & sources43.6% · raw 43.6 % Headline · unknown effort · Peer vendor Source/record date: 2026-09-03 As reported in OpenAI GPT-6 Astra table https://openai.com/index/gpt-6-astra/ |
| GDP.PDF | 35%unknown effort | All 1 recorded result & sources35% · raw 35 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 GDP.PDF all-pass rate https://deepmind.google/models/model-cards/gemini-3-8-flash |
| GDPval-AA v2 | 1545unknown effort | All 1 recorded result & sources1545 · raw 1545 Elo Headline · unknown effort · Own vendor Source/record date: 2026-09-03 Elo https://deepmind.google/models/model-cards/gemini-3-8-flash |
| GPQA Diamond | 95.3%unknown effort | All 1 recorded result & sources95.3% · raw 95.3 % Headline · unknown effort · Peer vendor Source/record date: 2026-09-03 As reported by OpenAI https://openai.com/index/gpt-6-astra/ |
| Harvey Legal Agent | 10%unknown effort | All 1 recorded result & sources10% · raw 10 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 All pass rate https://deepmind.google/models/model-cards/gemini-3-8-flash |
| HealthBench Professional | 52.1%unknown effort | All 1 recorded result & sources52.1% · raw 52.1 % Headline · unknown effort · Peer vendor Source/record date: 2026-09-03 As reported by OpenAI https://openai.com/index/gpt-6-astra/ |
| HLE-Verified | 54.9%unknown effort | All 1 recorded result & sources54.9% · raw 54.9 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 HLE-Verified https://deepmind.google/models/model-cards/gemini-3-8-flash |
| LABBench2 | 86.2%unknown effort | All 1 recorded result & sources86.2% · raw 86.2 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 LABBench2 https://deepmind.google/models/model-cards/gemini-3-8-flash |
| LVBench | 87.8%unknown effort | All 1 recorded result & sources87.8% · raw 87.8 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 LVBench long video (agentic) https://deepmind.google/models/model-cards/gemini-3-8-flash |
| OSWorld 2.0 | 59%unknown effort | All 1 recorded result & sources59% · raw 59 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 Partial score; batch tool enabled https://deepmind.google/models/model-cards/gemini-3-8-flash |
| Terminal-Bench 2.1 | 89.4%unknown effort | All 1 recorded result & sources89.4% · raw 89.4 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 https://deepmind.google/models/model-cards/gemini-3-8-flash |
| Terminal-Bench 4.0 | 19.1%unknown effort | All 1 recorded result & sources19.1% · raw 19.1 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 https://deepmind.google/models/model-cards/gemini-3-8-flash |
| Vals Finance Agent v2 | 61.4%unknown effort | All 1 recorded result & sources61.4% · raw 61.4 % Headline · unknown effort · Own vendor Source/record date: 2026-09-03 Vals Finance Agent v2 https://deepmind.google/models/model-cards/gemini-3-8-flash |
| AA Intelligence Index (vendor-cited) | 59unknown effort | All 1 recorded result & sources59 · raw 59 index Headline · unknown effort · Peer vendor Source/record date: Not recorded from chart/figure as reported on Ling-3.0-flash-VL HF card AA Index v4.1.1 chart (peer spillover); Gemini 3.8 Flash (high) https://huggingface.co/inclusionAI/Ling-3.0-flash-VL |
| Gray Swan IPI | 5.5%unknown effort | All 1 recorded result & sources5.5% · raw 5.5 % Headline · unknown effort · Peer vendor Source/record date: 2026-09-30 As reported by Google Gemini 4 Argon Gray Swan chart (K=15 attack success rate) https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini_4_cyber_evals_gray_swan_i.width-1200.format-webp.webp |
Independent evaluators
Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.
| Benchmark / evaluator | Headline score | Evidence |
|---|---|---|
| Intelligence Index · Artificial Analysis | 41high effort | All 3 recorded results & sources41 · raw 41 index Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; https://artificialanalysis.ai/models/gemini-3-8-flash40 · raw 40 index Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; https://artificialanalysis.ai/models/gemini-3-8-flash-medium33 · raw 33 index Alternative · low effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; https://artificialanalysis.ai/models/gemini-3-8-flash-low |
| Cost per Intelligence Index task · Artificial Analysis | $1.24high effort | All 2 recorded results & sources$1.24 · raw 1.24 USD Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; https://artificialanalysis.ai/models/gemini-3-8-flash$0.93 · raw 0.93 USD Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; https://artificialanalysis.ai/models/gemini-3-8-flash-medium |
| Output speed · Artificial Analysis | 249 tok/shigh effort | All 1 recorded result & sources249 tok/s · raw 249 tok/s Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Current AA model leaderboard; Median output tokens/s; leaderboard rounds to whole tokens. https://artificialanalysis.ai/models/gemini-3-8-flash |
| Vals Index · Vals AI | 54.8%unknown effort | All 1 recorded result & sources54.8% · raw 54.83 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-03 Refreshed from current Vals leaderboard. Cost/test $5.73. https://www.vals.ai/benchmarks/vals_index |
| Bugs fixed /105 · Bug Hunt Bench | 18 fixeshigh effort | All 1 recorded result & sources18 fixes · raw 18 fixes Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Harness: Antigravity CLI; effort high; 3 runs; evaluation 2026-09-15. Best documented score for this effort in Oct 1 README. Headline: best documented model run. https://github.com/phuryn/bug-hunt-bench |
| CUDA board % of roofline · KernelBench (community board) | 16.7%high effort | All 1 recorded result & sources16.7% · raw 16.7 % Headline · high effort · Independent evaluator Source/record date: 2026-09-23 Gemini 3.8 Flash (High) board; CUDA 3/4; Mega 2.74× https://kernelbench.com/models/gemini-3.8-flash-high |
| GDPval-AA Elo · Artificial Analysis | 1412high effort | All 1 recorded result & sources1412 · raw 1412 Elo Headline · high effort · Independent evaluator Source/record date: 2026-09-23 GDPval-AA v2.1 Elo; Gemini 3.8 Flash (high) https://artificialanalysis.ai/evaluations/gdpval-aa |
| ARC-AGI-1 · ARC Prize | 98.5%high effort | All 3 recorded results & sources98.5% · raw 98.5 % Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. Headline: best verified value; source matrix effort order breaks ties. https://arcprize.org/results/google-gemini-3-8-flash97.5% · raw 97.5 % Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. https://arcprize.org/results/google-gemini-3-8-flash90.5% · raw 90.5 % Alternative · low effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. https://arcprize.org/results/google-gemini-3-8-flash |
| ARC-AGI-2 · ARC Prize | 89.2%high effort | All 3 recorded results & sources89.2% · raw 89.2 % Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. Headline: best verified value; source matrix effort order breaks ties. https://arcprize.org/results/google-gemini-3-8-flash82.9% · raw 82.9 % Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. https://arcprize.org/results/google-gemini-3-8-flash77.5% · raw 77.5 % Alternative · low effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Published verified configuration. https://arcprize.org/results/google-gemini-3-8-flash |
| ARC-AGI-3 (Standard) · ARC Prize | 10.4%high effort | All 3 recorded results & sources10.4% · raw 10.37 % Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Standard (notes carry). Headline: best verified value; source matrix effort order breaks ties. https://arcprize.org/results/google-gemini-3-8-flash4% · raw 3.99 % Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Standard (notes carry). https://arcprize.org/results/google-gemini-3-8-flash6% · raw 5.99 % Alternative · low effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Standard (notes carry). https://arcprize.org/results/google-gemini-3-8-flash |
| ARC-AGI-3 · ARC Prize | 35%high effort | All 3 recorded results & sources35% · raw 35 % Headline · high effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Provider Adapter. Headline: best verified value; source matrix effort order breaks ties. https://arcprize.org/results/google-gemini-3-8-flash24.2% · raw 24.2 % Alternative · medium effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Provider Adapter. https://arcprize.org/results/google-gemini-3-8-flash15.1% · raw 15.07 % Alternative · low effort · Independent evaluator Source/record date: 2026-10-03 Semi-Private verified matrix. Harness: Provider Adapter. https://arcprize.org/results/google-gemini-3-8-flash |
| Money gain · Andon Labs | $4593.79unknown effort | All 1 recorded result & sources$4593.79 · raw 4593.79 $ Headline · unknown effort · Independent evaluator Source/record date: 2026-10-03 Vending-Bench 2 net gain = final_value in the page public vb2 data module minus $500 starting balance. Full 66-model source checked. https://andonlabs.com/evals/vending-bench-2 |
| Blueprint Bench · Andon Labs | 38.6%unknown effort | All 1 recorded result & sources38.6% · raw 38.6 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-03 Blueprint-Bench 2 connectivity similarity; published fractional score multiplied by 100. https://andonlabs.com/evals/blueprint-bench-2 |
| Average Score · WeirdML v3 | 7.4%high effort | All 1 recorded result & sources7.4% · raw 7.38 % Headline · high effort · Independent evaluator Source/record date: 2026-10-02 WeirdML variant Gemini 3.8 Flash (high); harness gemini_cli 0.59.0; values from prepared data JSON; raw 0.073787; official 80/20 aggregate (area 500k-50M tokens + final best) https://htihle.github.io/weirdml.html |
| Final Best Score · WeirdML v3 | 14.6%high effort | All 1 recorded result & sources14.6% · raw 14.57 % Headline · high effort · Independent evaluator Source/record date: 2026-10-02 WeirdML variant Gemini 3.8 Flash (high); harness gemini_cli 0.59.0; values from prepared data JSON; raw 0.145741; mean final best effective score https://htihle.github.io/weirdml.html |
| Cost / Run · WeirdML v3 | $3.94high effort | All 1 recorded result & sources$3.94 · raw 3.94 USD Headline · high effort · Independent evaluator Source/record date: 2026-10-02 WeirdML variant Gemini 3.8 Flash (high); harness gemini_cli 0.59.0; values from prepared data JSON; mean API cost per run, same task weighting as scores https://htihle.github.io/weirdml.html |
| Vibe Code Bench v1.1 · Vals AI | 78.7%unknown effort | All 1 recorded result & sources78.7% · raw 78.65 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-03 Refreshed from current Vals leaderboard. Harness: OpenHands. Cost/test $6.87. https://www.vals.ai/benchmarks/vibe-code |
Read how we select and source scores or the comparison guide.