Gemma 4 26B-A4B Benchmarks, Specifications & Availability

Gemma 4 26B-A4B

Explore Gemma 4 26B-A4B from Google: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Published specifications

Provider: Google
Access: Open weights
License: Apache-2.0
Context window: 256K
Total parameters: 25.2B
Active parameters: 3.8B
Released: 2026-03-31
Modalities: text, image, video
Family: Gemma 4

Hugging Face · Announcement · Website · OpenRouter

Model notes

Public ungated Apache-2.0 instruction-tuned checkpoint; PT, assistant/MTP and QAT builds are not extra curated models. Vendor architecture size is retained separately from HF tensor totals including modality modules. 26B-A4B is nominal; card publishes25.2B language parameters and3.8B active, plus a ~550M vision encoder. No native audio input; video uses frames (up to60 seconds at1fps), text output only. Context is the published configured limit, not a universal guarantee of whole-window accuracy. Reasoning is an on/off mode, not an API effort level. Report Table9 disables thinking; vision uses1120 or280 visual-token budgets. OmniDocBench uses native average edit distance, not accuracy. Card/report audio aggregates and the older launch τ² figures differ and remain source-specific observations. Official release notes date31March; evaluator release dates may say2April and do not override the vendor date. Hosted endpoint precision/provider conditions are separate from reference-weight benchmark evidence.

Compare Gemma 4 26B-A4B with other models →Compare subscription plans Explore 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.

Recorded pricing tiers
Tier	Input / 1M	Output / 1M	Cached input / 1M	Cache write / 1M	Date & source
Default	$0.042	$0.22	$0.021	—	2026-10-20 · OpenRouter source

Recorded pricing notes

Min-healthy endpoint minima checked19October2026: input $0.042/1M from Darkbloom (darkbloom, 100% 30m uptime); output $0.18/1M from Makora (makora, 100% 30m uptime); cached $0.021/1M from Darkbloom (darkbloom, 100% 30m uptime). Each metric may name a different provider; these are not one combined endpoint quote. Published discounted rates are already inclusive; flags are not applied a second time. Hosted precision/service tiers do not establish reference BF16 benchmark precision. No contributor sibling found in all53-model audit. Keyless API throughput is unavailable, not zero.; min-healthy endpoint minima 2026-10-19; min-healthy endpoint minima 2026-10-20

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
MMLU-Pro	82.6%unknown effort	All 2 recorded results & sources 82.6% · raw 82.6 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5 85.2% · raw 85.2 % Alternative · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
AIME 2026	88.3%unknown effort	All 1 recorded result & sources 88.3% · raw 88.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. No tools. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
LiveCodeBench v6	77.1%unknown effort	All 1 recorded result & sources 77.1% · raw 77.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Codeforces Rating	1718unknown effort	All 1 recorded result & sources 1718 · raw 1718 rating Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
SciCode (Gemma report; aggregation unspecified)	40%unknown effort	All 1 recorded result & sources 40% · raw 40 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
GPQA Diamond	82.3%unknown effort	All 2 recorded results & sources 82.3% · raw 82.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5 79.6% · raw 79.61 % Alternative · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
BigBench Extra Hard (micro average)	64.8%unknown effort	All 1 recorded result & sources 64.8% · raw 64.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Humanity's Last Exam (no tools)	8.7%unknown effort	All 1 recorded result & sources 8.7% · raw 8.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Humanity's Last Exam (with tools)	17.2%unknown effort	All 1 recorded result & sources 17.2% · raw 17.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Search tool enabled. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
IFBench (Gemma report protocol)	72%unknown effort	All 1 recorded result & sources 72% · raw 72 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. IFBench strict/loose convention is not stated in this table. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
IFEval	98.5%unknown effort	All 1 recorded result & sources 98.5% · raw 98.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MMMLU	86.3%unknown effort	All 1 recorded result & sources 86.3% · raw 86.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MRCR v2 (8 needles,128K average)	44.1%unknown effort	All 1 recorded result & sources 44.1% · raw 44.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Terminal Bench Hard (Gemma report)	14%unknown effort	All 1 recorded result & sources 14% · raw 14 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Airline)	76%unknown effort	All 1 recorded result & sources 76% · raw 76 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Retail)	85.5%unknown effort	All 1 recorded result & sources 85.5% · raw 85.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Telecom)	43%unknown effort	All 1 recorded result & sources 43% · raw 43 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MMMU Pro (no tools)	73.8%unknown effort	All 1 recorded result & sources 73.8% · raw 73.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MMMU Pro (Gemma,280 visual tokens)	73.2%unknown effort	All 1 recorded result & sources 73.2% · raw 73.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
MATH-Vision	82.4%unknown effort	All 1 recorded result & sources 82.4% · raw 82.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MATH-Vision (Gemma,280 visual tokens)	80.3%unknown effort	All 1 recorded result & sources 80.3% · raw 80.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
MedXpertQA MM	58.1%unknown effort	All 1 recorded result & sources 58.1% · raw 58.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MedXpertQA MM (Gemma,280 visual tokens)	55.7%unknown effort	All 1 recorded result & sources 55.7% · raw 55.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
InfographicVQA	89.3%unknown effort	All 1 recorded result & sources 89.3% · raw 89.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
InfographicVQA (Gemma,280 visual tokens)	77.8%unknown effort	All 1 recorded result & sources 77.8% · raw 77.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
OmniDocBench1.5 (average edit distance)	0.149 edit distanceunknown effort	All 1 recorded result & sources 0.149 edit distance · raw 0.149 edit distance Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native edit distance. https://arxiv.org/pdf/2607.02770#page=6
OmniDocBench1.5 (average edit distance) (Gemma,280 visual tokens)	0.269 edit distanceunknown effort	All 1 recorded result & sources 0.269 edit distance · raw 0.269 edit distance Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
RULER32K (Gemma,no thinking)	97.3%none effort	All 1 recorded result & sources 97.3% · raw 97.3 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. RULER32K. Native %. https://arxiv.org/pdf/2607.02770#page=8
RULER128K (Gemma,no thinking)	89.8%none effort	All 1 recorded result & sources 89.8% · raw 89.8 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. RULER128K. Native %. https://arxiv.org/pdf/2607.02770#page=8
LOFT text retrieval128K (Recall@k) (Gemma,no thinking)	66.3%none effort	All 1 recorded result & sources 66.3% · raw 66.3 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. LOFT text retrieval128K (Recall@k). Native %. https://arxiv.org/pdf/2607.02770#page=8
GraphWalks <128K (F1) (Gemma,no thinking)	72.6%none effort	All 1 recorded result & sources 72.6% · raw 72.6 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. GraphWalks <128K (F1). Native %. https://arxiv.org/pdf/2607.02770#page=8
MTOB English→Kalamang ~128K half-book (Gemma,no thinking)	50 chrFnone effort	All 1 recorded result & sources 50 chrF · raw 50 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB English→Kalamang ~128K half-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB English→Kalamang ~256K full-book (Gemma,no thinking)	48.9 chrFnone effort	All 1 recorded result & sources 48.9 chrF · raw 48.9 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB English→Kalamang ~256K full-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB Kalamang→English ~128K half-book (Gemma,no thinking)	45 chrFnone effort	All 1 recorded result & sources 45 chrF · raw 45 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB Kalamang→English ~128K half-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB Kalamang→English ~256K full-book (Gemma,no thinking)	42.7 chrFnone effort	All 1 recorded result & sources 42.7 chrF · raw 42.7 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB Kalamang→English ~256K full-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
τ²-bench (Tau-2)	68.2%unknown effort	All 1 recorded result & sources 68.2% · raw 68.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google official card; published average over3 domains, thinking on. Not recomputed from report domains. https://huggingface.co/google/gemma-4-26B-A4B-it/blob/4d7ae4984b7db7de8f8457170b3f1a419ee76d52/README.md
Arena Text (Google-cited19June2026)	1438unknown effort	All 1 recorded result & sources 1438 · raw 1438 Elo Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table4; human Arena ratings as of19June2026,95%CI±8; model execution dates/quantization unpublished. https://arxiv.org/pdf/2607.02770#page=4
Arena Text (Google-cited2April2026)	1441unknown effort	All 1 recorded result & sources 1441 · raw 1441 Elo Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google launch figure; source says as of4/2/26. Visually checked exact printed value; no estimate from unlabeled plot. https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/
MMMLU (Gemma launch figure label)	82.6%unknown effort	All 1 recorded result & sources 82.6% · raw 82.6 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: literal Google launch figure label; benchmark-name conflict with full card/report; this cell is not used as ordinary MMLU-Pro or MMMLU. https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/
AA-Omniscience (NVIDIA Lightning protocol)	22.2 scoreunknown effort	All 1 recorded result & sources 22.2 score · raw 22.17 score Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. The broad source label does not establish the signed AA index versus accuracy quantity; preserve the native score, without ranking or substituting it for directly labeled report components. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
Humanity's Last Exam (text-only, no tools)	17.4%unknown effort	All 1 recorded result & sources 17.4% · raw 17.42 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
SciCode (NVIDIA Lightning protocol)	40.3%unknown effort	All 1 recorded result & sources 40.3% · raw 40.28 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
SWE-bench Verified (NVIDIA Lightning protocol)	57.4%unknown effort	All 1 recorded result & sources 57.4% · raw 57.4 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
SWE-bench Multilingual (NVIDIA Lightning protocol)	43.4%unknown effort	All 1 recorded result & sources 43.4% · raw 43.4 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
Terminal-Bench 2.1 (NVIDIA Lightning protocol)	37.2%unknown effort	All 1 recorded result & sources 37.2% · raw 37.22 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
PinchBench	74.7%unknown effort	All 1 recorded result & sources 74.7% · raw 74.7 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
BrowseComp (NVIDIA Lightning protocol)	26.3%unknown effort	All 1 recorded result & sources 26.3% · raw 26.3 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
τ³-bench Banking	14%unknown effort	All 1 recorded result & sources 14% · raw 14.02 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
GDPval-AA v2 (Lightning release committee Elo)	807unknown effort	All 1 recorded result & sources 807 · raw 807 Elo Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
IFBench loose (Lightning release)	77.3%unknown effort	All 1 recorded result & sources 77.3% · raw 77.25 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md
AA-LCR	57.6%unknown effort	All 1 recorded result & sources 57.6% · raw 57.56 % Headline · unknown effort · Peer vendor Source/record date: 2026-10-19 Harness: NVIDIA same-suite vendor comparison of exact Gemma4 26B-A4B-it (confirmed figure label). As reported by NVIDIA; effort and inference precision unpublished for the peer. https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

Independent evaluator headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
GDPval-AA v2.1 Elo · Artificial Analysis	566unknown effort	All 1 recorded result & sources 566 · raw 566 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-20 Manually observed public GDPval-AA v2.1 leaderboard table, full 298-configuration pool checked 20 October 2026. Exact source configuration: Gemma 4 26B A4B (Reasoning); creator: Google; source rank 216; source release-month label Apr 2026 is model release metadata, not execution date. Published rounded native Elo 566; literal displayed CI -20 / +20. Native Elo is not a percentage and is not the normalized (Elo−500)/2000 chart quantity. This versioned metric is separate from older generic GDPval-AA Elo and vendor GDPval/GDPval-AA results; no old evidence is overwritten. Execution date, inference precision, unprinted effort budgets and task costs are unpublished in this captured table. Elo is relative to the dated comparison pool; drift from an earlier observation does not prove a new evaluation run. Source says Reasoning (on), not an API effort level; reasoning effort remains unknown. Headline: highest published native Elo for this exact approved model; tied values prefer existing featured effort then source order. https://artificialanalysis.ai/evaluations/gdpval-aa
Intelligence Index (estimated) · Artificial Analysis	16.7unknown effort	All 2 recorded results & sources 16.7 · raw 16.671930040086 index Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 13.1 · raw 13.1303772801787 index Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
AA-Omniscience Index · Artificial Analysis	-50.8unknown effort	All 2 recorded results & sources -50.8 · raw -50.833333333333336 index Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b -62.3 · raw -62.31666666666667 index Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
AA-Omniscience accuracy · Artificial Analysis	19.1%unknown effort	All 2 recorded results & sources 19.1% · raw 19.1 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 15.5% · raw 15.483333333333333 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
AA-Omniscience hallucination rate · Artificial Analysis	86.4%unknown effort	All 2 recorded results & sources 86.4% · raw 86.44416975690152 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 92.1% · raw 92.05284953658057 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
GDPval-AA Elo · Artificial Analysis	565.8unknown effort	All 1 recorded result & sources 565.8 · raw 565.75 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
Terminal-Bench 2.1 (AA) · Artificial Analysis	39%unknown effort	All 1 recorded result & sources 39% · raw 38.9513108614232 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
Terminal Bench Hard (AA) · Artificial Analysis	13.6%unknown effort	All 2 recorded results & sources 13.6% · raw 13.639999999999999 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 25% · raw 25 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
SciCode (AA) · Artificial Analysis	40%unknown effort	All 1 recorded result & sources 40% · raw 40.046296296296305 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA-LCR · Artificial Analysis	65.7%unknown effort	All 2 recorded results & sources 65.7% · raw 65.6666666666667 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 42.3% · raw 42.3333333333333 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
IFBench (AA) · Artificial Analysis	72.4%unknown effort	All 2 recorded results & sources 72.4% · raw 72.4489795918367 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 45.4% · raw 45.442176870748305 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
Humanity's Last Exam (AA harness) · Artificial Analysis	19.3%unknown effort	All 2 recorded results & sources 19.3% · raw 19.3234476367006 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 11.5% · raw 11.4921223354958 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
GPQA Diamond (AA) · Artificial Analysis	79.2%unknown effort	All 2 recorded results & sources 79.2% · raw 79.1919191919192 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 71.4% · raw 71.4141414141414 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
CritPt (AA) · Artificial Analysis	0%unknown effort	All 2 recorded results & sources 0% · raw 0 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 0% · raw 0 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
AutomationBench (AA) · Artificial Analysis	1.9%unknown effort	All 1 recorded result & sources 1.9% · raw 1.9430215909597492 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
MMMU-Pro (AA) · Artificial Analysis	69.2%unknown effort	All 2 recorded results & sources 69.2% · raw 69.2485549132948 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 66.7% · raw 66.7052023121387 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
τ²-Telecom (AA) · Artificial Analysis	43.6%unknown effort	All 2 recorded results & sources 43.6% · raw 43.5672514619883 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 40.4% · raw 40.3508771929825 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning
τ³-Banking (AA) · Artificial Analysis	12%unknown effort	All 1 recorded result & sources 12% · raw 11.9587628865979 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
ITBench-AA · Artificial Analysis	23.6%unknown effort	All 1 recorded result & sources 23.6% · raw 23.6346516007533 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA Industry Index · Economics · Artificial Analysis	18.9unknown effort	All 1 recorded result & sources 18.9 · raw 18.9354150061785 index Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA-Briefcase Elo · Artificial Analysis	315.1unknown effort	All 1 recorded result & sources 315.1 · raw 315.08 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA-Briefcase analytical quality Elo · Artificial Analysis	29.1unknown effort	All 1 recorded result & sources 29.1 · raw 29.11 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA-Briefcase presentation Elo · Artificial Analysis	275.5unknown effort	All 1 recorded result & sources 275.5 · raw 275.54 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
AA-Briefcase rubric pass rate · Artificial Analysis	7.3%unknown effort	All 1 recorded result & sources 7.3% · raw 7.329842931937172 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b
Output speed · Artificial Analysis	69.5 tok/sunknown effort	All 2 recorded results & sources 69.5 tok/s · raw 69.5344132390526 tok/s Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b 99.4 tok/s · raw 99.3806723551139 tok/s Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-26B-A4B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-26b-a4b-non-reasoning

Read how we select and source scores or the comparison guide.