Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Gemma 4 E4B
~47
0/8 categoriesGLM-5
75
Winner · 4/8 categoriesGemma 4 E4B· GLM-5
Pick GLM-5 if you want the stronger benchmark profile. Gemma 4 E4B only becomes the better choice if you want the stronger reasoning-first profile.
GLM-5 is clearly ahead on the aggregate, 75 to 47. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
GLM-5's sharpest advantage is in reasoning, where it averages 70.5 against 25.4. The single biggest benchmark swing on the page is BBH, 33.1% to 83%.
Gemma 4 E4B is the reasoning model in the pair, while GLM-5 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. GLM-5 gives you the larger context window at 200K, compared with 128K for Gemma 4 E4B.
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 E4B | GLM-5 |
|---|---|---|
| Agentic | ||
| Terminal-Bench 2.0 | — | 56.2% |
| BrowseComp | — | 62% |
| OSWorld-Verified | — | 58% |
| Claw-Eval | — | 57.7% |
| QwenClawBench | — | 54.1% |
| QwenWebBench | — | 1315 |
| TAU3-Bench | — | 65.6% |
| VITA-Bench | — | 37.0% |
| DeepPlanning | — | 14.6% |
| Toolathlon | — | 38% |
| MCP Atlas | — | 31.1% |
| MCP-Tasks | — | 60.8% |
| WideResearch | — | 69.8% |
| Tau2-Airline | — | 80.5% |
| Tau2-Telecom | — | 98.2% |
| PinchBench | — | 86.4% |
| BFCL v4 | — | 70.8% |
| CodingGLM-5 wins | ||
| LiveCodeBench | 52% | 52% |
| HumanEval | — | 80% |
| SWE-bench Verified | — | 77.8% |
| SWE-bench Verified* | — | 72.8% |
| LiveCodeBench v6 | — | 85.6% |
| SWE-bench Pro | — | 55.1% |
| SWE Multilingual | — | 73.3% |
| NL2Repo | — | 35.9% |
| SWE-Rebench | — | 62.8% |
| React Native Evals | — | 74.2% |
| Multimodal & GroundedGLM-5 wins | ||
| MMMU-Pro | 52.6% | 66% |
| OfficeQA Pro | — | 73% |
| ReasoningGLM-5 wins | ||
| BBH | 33.1% | 83% |
| MRCRv2 | 25.4% | 73% |
| MuSR | — | 82% |
| LongBench v2 | — | 60.8% |
| AI-Needle | — | 63.3% |
| KnowledgeGLM-5 wins | ||
| GPQA | 58.6% | 86% |
| MMLU-Pro | 69.4% | 85.7% |
| MMLU | — | 91.7% |
| GPQA-D | — | 81.6% |
| SuperGPQA | — | 66.8% |
| MMLU-Pro (Arcee) | — | 85.8% |
| MMLU-Redux | — | 94.4% |
| C-Eval | — | 92.8% |
| HLE | — | 27.2% |
| FrontierScience | — | 74% |
| SimpleQA | — | 84% |
| Instruction Following | ||
| IFEval | — | 92.6% |
| IFBench | — | 72.3% |
| Multilingual | ||
| MGSM | — | 84% |
| MMLU-ProX | — | 83.1% |
| NOVA-63 | — | 55.1% |
| INCLUDE | — | 84.9% |
| PolyMath | — | 65.2% |
| VWT2k-lite | — | 82.1% |
| MAXIFE | — | 85.6% |
| Mathematics | ||
| AIME 2023 | — | 88% |
| AIME 2024 | — | 90% |
| AIME 2025 | — | 93.3% |
| AIME26 | — | 95.8% |
| AIME25 (Arcee) | — | 93.3% |
| HMMT Feb 2023 | — | 84% |
| HMMT Feb 2024 | — | 86% |
| HMMT Feb 2025 | — | 85% |
| HMMT Feb 2025 | — | 97.5% |
| HMMT Nov 2025 | — | 96.9% |
| HMMT Feb 2026 | — | 86.4% |
| MMAnswerBench | — | 82.5% |
| BRUMO 2025 | — | 87% |
| MATH-500 | — | 97.4% |
GLM-5 is ahead overall, 75 to 47. The biggest single separator in this matchup is BBH, where the scores are 33.1% and 83%.
GLM-5 has the edge for knowledge tasks in this comparison, averaging 67.7 versus 65.6. Inside this category, GPQA is the benchmark that creates the most daylight between them.
GLM-5 has the edge for coding in this comparison, averaging 60.4 versus 52. Gemma 4 E4B stays close enough that the answer can still flip depending on your workload.
GLM-5 has the edge for reasoning in this comparison, averaging 70.5 versus 25.4. Inside this category, BBH is the benchmark that creates the most daylight between them.
GLM-5 has the edge for multimodal and grounded tasks in this comparison, averaging 69.2 versus 52.6. 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.