Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Gemma 4 E2B
~39
Winner · 3/8 categoriesGLM-4.5-Air
38
1/8 categoriesGemma 4 E2B· GLM-4.5-Air
Pick Gemma 4 E2B if you want the stronger benchmark profile. GLM-4.5-Air only becomes the better choice if reasoning is the priority or you would rather avoid the extra latency and token burn of a reasoning model.
Gemma 4 E2B finishes one point ahead overall, 39 to 38. That is enough to call, but not enough to treat as a blowout. This matchup comes down to a few meaningful edges rather than one model dominating the board.
Gemma 4 E2B's sharpest advantage is in knowledge, where it averages 54.1 against 31. The single biggest benchmark swing on the page is BBH, 21.9% to 63%. GLM-4.5-Air does hit back in reasoning, so the answer changes if that is the part of the workload you care about most.
Gemma 4 E2B is the reasoning model in the pair, while GLM-4.5-Air 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 E2B | GLM-4.5-Air |
|---|---|---|
| Agentic | ||
| Terminal-Bench 2.0 | — | 28% |
| BrowseComp | — | 37% |
| Claw-Eval | — | 42.3% |
| CodingGemma 4 E2B wins | ||
| LiveCodeBench | 44% | 15% |
| HumanEval | — | 27% |
| SWE-bench Verified | — | 15% |
| SWE-bench Pro | — | 14% |
| SWE-Rebench | — | 38.3% |
| Multimodal & GroundedGemma 4 E2B wins | ||
| MMMU-Pro | 44.2% | 36% |
| ReasoningGLM-4.5-Air wins | ||
| BBH | 21.9% | 63% |
| MRCRv2 | 19.1% | 51% |
| MuSR | — | 31% |
| LongBench v2 | — | 47% |
| KnowledgeGemma 4 E2B wins | ||
| GPQA | 43.4% | 34% |
| MMLU-Pro | 60% | 51% |
| MMLU | — | 35% |
| SuperGPQA | — | 32% |
| HLE | — | 4% |
| FrontierScience | — | 37% |
| SimpleQA | — | 33% |
| Instruction Following | ||
| IFEval | — | 68% |
| Multilingual | ||
| MGSM | — | 63% |
| MMLU-ProX | — | 57% |
| Mathematics | ||
| AIME 2023 | — | 35% |
| AIME 2024 | — | 37% |
| AIME 2025 | — | 36% |
| HMMT Feb 2023 | — | 31% |
| HMMT Feb 2024 | — | 33% |
| HMMT Feb 2025 | — | 32% |
| BRUMO 2025 | — | 34% |
| MATH-500 | — | 57% |
Gemma 4 E2B is ahead overall, 39 to 38. The biggest single separator in this matchup is BBH, where the scores are 21.9% and 63%.
Gemma 4 E2B has the edge for knowledge tasks in this comparison, averaging 54.1 versus 31. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Gemma 4 E2B has the edge for coding in this comparison, averaging 44 versus 22.9. Inside this category, LiveCodeBench is the benchmark that creates the most daylight between them.
GLM-4.5-Air has the edge for reasoning in this comparison, averaging 44.1 versus 19.1. Inside this category, BBH is the benchmark that creates the most daylight between them.
Gemma 4 E2B has the edge for multimodal and grounded tasks in this comparison, averaging 44.2 versus 36. Inside this category, MMMU-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.