Across 11 shared benchmarks, Gemma 3 1B scores higher on 0 and Gemma 3 27B on 11. The widest gap is MMLU-Pro, where Gemma 3 27B scores 67.6 against 14.7.
| Benchmark | Gemma 3 1B | Gemma 3 27B |
|---|---|---|
| bbh | 39.1 | 87.6 |
| BIG-Bench Extra Hard | 7.2 | 19.3 |
| Global-MMLU-Lite | 34.2 | 75.1 |
| GPQA Diamond | 19.2 | 46 |
| GSM8K | 62.8 | 95.9 |
| humaneval | 41.5 | 87.8 |
| ifeval | 80.2 | 90.4 |
| LiveCodeBench | 1.9 | 29.7 |
| MBPP | 0.4 | 74.4 |
| MMLU-Pro | 14.7 | 67.6 |
| simpleqa | 2.2 | 10 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.