Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Gemma 4 26B A4B
64
Winner · 0/8 categoriesMistral Small 4
~62
1/8 categoriesGemma 4 26B A4B· Mistral Small 4
Pick Gemma 4 26B A4B if you want the stronger benchmark profile. Mistral Small 4 only becomes the better choice if knowledge is the priority or you would rather avoid the extra latency and token burn of a reasoning model.
Gemma 4 26B A4B has the cleaner overall profile here, landing at 64 versus 62. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Gemma 4 26B A4B is the reasoning model in the pair, while Mistral Small 4 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use.
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | Gemma 4 26B A4B | Mistral Small 4 |
|---|---|---|
| Agentic | ||
| Coming soon | ||
| Coding | ||
| LiveCodeBench | 77.1% | — |
| HumanEval | — | 84.8% |
| Multimodal & Grounded | ||
| MMMU-Pro | 73.8% | — |
| Reasoning | ||
| BBH | 64.8% | — |
| MRCRv2 | 44.1% | — |
| KnowledgeMistral Small 4 wins | ||
| GPQA | 82.3% | 45.3% |
| MMLU-Pro | 82.6% | 66.3% |
| HLE | 17.2% | — |
| HLE w/o tools | 8.7% | — |
| Instruction Following | ||
| IFEval | — | 82.9% |
| Multilingual | ||
| Coming soon | ||
| Mathematics | ||
| Coming soon | ||
Gemma 4 26B A4B is ahead overall, 64 to 62. The biggest single separator in this matchup is GPQA, where the scores are 82.3% and 45.3%.
Mistral Small 4 has the edge for knowledge tasks in this comparison, averaging 58.9 versus 56.1. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.