Gemma 4 12B.
Google DeepMind Jun 2026 Open weights Small
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BBEH micro average | 53% | #5 of 15 | Lab model card | |
| EQ-Bench 3 | 1,416 | #32 of 78 | Official leaderboard | |
| AgentIF (2026-10 setting) | 63.2% | #10 of 24 | Official leaderboard | |
| SEA-HELM | 69.3 | #34 of 71 | Official leaderboard | |
| MMMLU | 83.4% | #37 of 71 | Lab model card | |
| MMLU-Pro | 77.2% | #99 of 167 | Lab model card | |
| Creative Writing v3 | 61.9 | #79 of 120 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 48.7 | #85 of 119 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 191.0 | #62 of 77 | Official leaderboard | |
| MRCR v2 | 43.4% | #25 of 30 | Lab model card | |
| Humanity's Last Exam | 5.20% | #152 of 154 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 78.8% | #126 of 302 | Lab model card | |
| AIME 2026 | 77.5% | #33 of 38 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 72% | #52 of 90 | Lab model card | |
| Codeforces rating | 1,659 | #39 of 47 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| tau2-bench | 69% | #59 of 79 | Lab model card | |
| AdaExploreBench | 36.5% | #6 of 8 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MedXpertQA-MM | 48.7% | #14 of 21 | Lab model card | |
| TRLawBench | 7.82 | #29 of 34 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CoVoST2 | 38.5 | #6 of 14 | Lab model card | |
| MATH-Vision (MathVision) | 79.7% | #22 of 51 | Lab model card | |
| MMMU-Pro | 69.1% | #53 of 90 | Lab model card | |
| OmniDocBench v1.5 | 0.16 | #8 of 9 | Lab model card | |
| ExtractBench | 45.4% | #26 of 27 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 59.7 | #140 of 185 | Official leaderboard |
The Self-Reported Index fits one score to the 12 capability results Google DeepMind reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 135.1 to 139.9. How we compute it.
The benchmarks the lab chose to headline, with the scores it gave.