Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Granite-4.0-H-350M
~24
0/8 categoriesLFM2.5-350M
~39
Winner · 2/8 categoriesGranite-4.0-H-350M· LFM2.5-350M
Pick LFM2.5-350M if you want the stronger benchmark profile. Granite-4.0-H-350M only becomes the better choice if its workflow or ecosystem matters more than the raw scoreboard.
LFM2.5-350M is clearly ahead on the aggregate, 39 to 24. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
LFM2.5-350M's sharpest advantage is in instruction following, where it averages 77 against 55.4. The single biggest benchmark swing on the page is IFEval, 55.4% to 77.0%.
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-H-350M | LFM2.5-350M |
|---|---|---|
| Agentic | ||
| Coming soon | ||
| Coding | ||
| HumanEval | 39% | — |
| Multimodal & Grounded | ||
| Coming soon | ||
| Reasoning | ||
| BBH | 33.1% | — |
| KnowledgeLFM2.5-350M wins | ||
| MMLU | 35.0% | — |
| GPQA | 24.1% | 30.6% |
| MMLU-Pro | 12.1% | 20.0% |
| Instruction FollowingLFM2.5-350M wins | ||
| IFEval | 55.4% | 77.0% |
| Multilingual | ||
| MGSM | 14.7% | — |
| Mathematics | ||
| Coming soon | ||
LFM2.5-350M is ahead overall, 39 to 24. The biggest single separator in this matchup is IFEval, where the scores are 55.4% and 77.0%.
LFM2.5-350M has the edge for knowledge tasks in this comparison, averaging 23.8 versus 16.4. Inside this category, MMLU-Pro is the benchmark that creates the most daylight between them.
LFM2.5-350M has the edge for instruction following in this comparison, averaging 77 versus 55.4. Inside this category, IFEval 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.