Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Granite-4.0-350M
~27
0/8 categoriesMistral 7B v0.3
~29
Winner · 1/8 categoriesGranite-4.0-350M· Mistral 7B v0.3
Pick Mistral 7B v0.3 if you want the stronger benchmark profile. Granite-4.0-350M only becomes the better choice if its workflow or ecosystem matters more than the raw scoreboard.
Mistral 7B v0.3 has the cleaner overall profile here, landing at 29 versus 27. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Mistral 7B v0.3's sharpest advantage is in knowledge, where it averages 31.5 against 18.5. The single biggest benchmark swing on the page is MMLU, 36.2% to 60.1%.
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 | Granite-4.0-350M | Mistral 7B v0.3 |
|---|---|---|
| Agentic | ||
| Coming soon | ||
| Coding | ||
| HumanEval | 38% | 30.5% |
| Multimodal & Grounded | ||
| Coming soon | ||
| Reasoning | ||
| BBH | 33.3% | — |
| KnowledgeMistral 7B v0.3 wins | ||
| MMLU | 36.2% | 60.1% |
| GPQA | 26.1% | — |
| MMLU-Pro | 14.4% | — |
| FrontierScience | — | 34% |
| SimpleQA | — | 28% |
| Instruction Following | ||
| IFEval | 61.6% | — |
| Multilingual | ||
| MGSM | 16.2% | — |
| Mathematics | ||
| Coming soon | ||
Mistral 7B v0.3 is ahead overall, 29 to 27. The biggest single separator in this matchup is MMLU, where the scores are 36.2% and 60.1%.
Mistral 7B v0.3 has the edge for knowledge tasks in this comparison, averaging 31.5 versus 18.5. Inside this category, MMLU 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.