Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Gemma 4 E2B
~39
1/8 categoriesLFM2.5-350M
~39
0/8 categoriesGemma 4 E2B· LFM2.5-350M
Treat this as a split decision. Gemma 4 E2B makes more sense if knowledge is the priority or you need the larger 128K context window; LFM2.5-350M is the better fit if you would rather avoid the extra latency and token burn of a reasoning model.
Gemma 4 E2B and LFM2.5-350M finish on the same overall score, so this is less about a single winner and more about where the edge shows up. The headline says tie; the benchmark table is where the real choice happens.
Gemma 4 E2B is the reasoning model in the pair, while LFM2.5-350M 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. Gemma 4 E2B gives you the larger context window at 128K, compared with 32K for LFM2.5-350M.
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 E2B | LFM2.5-350M |
|---|---|---|
| Agentic | ||
| Coming soon | ||
| Coding | ||
| LiveCodeBench | 44% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 44.2% | — |
| Reasoning | ||
| BBH | 21.9% | — |
| MRCRv2 | 19.1% | — |
| KnowledgeGemma 4 E2B wins | ||
| GPQA | 43.4% | 30.6% |
| MMLU-Pro | 60% | 20.0% |
| Instruction Following | ||
| IFEval | — | 77.0% |
| Multilingual | ||
| Coming soon | ||
| Mathematics | ||
| Coming soon | ||
Gemma 4 E2B and LFM2.5-350M are tied on overall score, so the right pick depends on which category matters most for your use case.
Gemma 4 E2B has the edge for knowledge tasks in this comparison, averaging 54.1 versus 23.8. Inside this category, MMLU-Pro 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.