Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
GLM-5
75
0/8 categoriesHolo3-35B-A3B
~78
Winner · 1/8 categoriesGLM-5· Holo3-35B-A3B
Pick Holo3-35B-A3B if you want the stronger benchmark profile. GLM-5 only becomes the better choice if you want the cheaper token bill or you need the larger 200K context window.
Holo3-35B-A3B has the cleaner overall profile here, landing at 78 versus 75. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Holo3-35B-A3B's sharpest advantage is in agentic, where it averages 77.8 against 58.3. The single biggest benchmark swing on the page is OSWorld-Verified, 58% to 77.8%.
Holo3-35B-A3B is also the more expensive model on tokens at $0.25 input / $1.80 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for GLM-5. That is roughly Infinityx on output cost alone. GLM-5 gives you the larger context window at 200K, compared with 64K for Holo3-35B-A3B.
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 | GLM-5 | Holo3-35B-A3B |
|---|---|---|
| AgenticHolo3-35B-A3B wins | ||
| Terminal-Bench 2.0 | 56.2% | — |
| BrowseComp | 62% | — |
| OSWorld-Verified | 58% | 77.8% |
| Coding | ||
| HumanEval | 80% | — |
| SWE-bench Verified | 77.8% | — |
| LiveCodeBench | 52% | — |
| SWE-bench Pro | 46% | — |
| SWE-Rebench | 62.8% | — |
| React Native Evals | 74.2% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 66% | — |
| OfficeQA Pro | 73% | — |
| Reasoning | ||
| MuSR | 82% | — |
| BBH | 83% | — |
| LongBench v2 | 77% | — |
| MRCRv2 | 73% | — |
| Knowledge | ||
| MMLU | 91.7% | — |
| GPQA | 86% | — |
| SuperGPQA | 84% | — |
| MMLU-Pro | 82% | — |
| HLE | 30.5% | — |
| FrontierScience | 74% | — |
| SimpleQA | 84% | — |
| Instruction Following | ||
| IFEval | 85% | — |
| Multilingual | ||
| MGSM | 84% | — |
| MMLU-ProX | 81% | — |
| Mathematics | ||
| AIME 2023 | 88% | — |
| AIME 2024 | 90% | — |
| AIME 2025 | 93.3% | — |
| HMMT Feb 2023 | 84% | — |
| HMMT Feb 2024 | 86% | — |
| HMMT Feb 2025 | 85% | — |
| BRUMO 2025 | 87% | — |
| MATH-500 | 97.4% | — |
Holo3-35B-A3B is ahead overall, 78 to 75. The biggest single separator in this matchup is OSWorld-Verified, where the scores are 58% and 77.8%.
Holo3-35B-A3B has the edge for agentic tasks in this comparison, averaging 77.8 versus 58.3. Inside this category, OSWorld-Verified 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.