GSM8K#8 of 28
exact_match
View run →86.3%±0.9%
Evaluated across 13 benchmarks. Ranks in the top 3 on 2 of 13. Strongest showing on GSM8K (86.3% exact_match, #8 of 28). Weakest on GPQA Extended (33.3% acc, #7 of 28).
| Benchmark | Metric | Score | Rank | Actions |
|---|---|---|---|---|
| GSM8K | exact_match | 86.3%±0.9% | #8 of 28 | View run → |
| IFEval | prompt_level_strict_acc | 81.0%±1.7% | #12 of 28 | View run → |
| EQ-Bench | eqbench | 72.7±1.9 | #21 of 28 | View run → |
| MBPP(Instruct) | pass@1 | 70.8%±2.0% | #2 of 24 | View run → |
| MMLU | acc | 70.7%±0.4% | #13 of 28 | View run → |
| MGSM | exact_match | 68.4%±0.8% | #10 of 28 | View run → |
| MBPP | pass@1 | 65.4%±2.1% | #9 of 28 | View run → |
| BBH | exact_match | 62.3%±0.4% | #6 of 28 | View run → |
| MMLU-Pro | exact_match | 59.7%±0.4% | #15 of 28 | View run → |
| LongBench | aggregate | 46.2%±0.5% | #1 of 17 | View run → |
| GPQA Diamond | acc | 35.9%±3.4% | #7 of 28 | View run → |
| GPQA Main | acc | 35.7%±2.3% | #6 of 28 | View run → |
| GPQA Extended | acc | 33.3%±2.0% | #7 of 28 | View run → |
Citation
FrozeBench. "google/gemma-3-12b-it." https://frozebench.com/models/google-gemma-3-12b-it. Retrieved 2026-06-04.
BibTeX
@misc{frozebench_google_gemma_3_12b_it,
title = {google/gemma-3-12b-it},
howpublished = {\url{https://frozebench.com/models/google-gemma-3-12b-it}},
year = {2026},
note = {FrozeBench. Retrieved 2026-06-04.}
}URL
https://frozebench.com/models/google-gemma-3-12b-it