Comparing 2 models
Gemini 3.1 Pro Preview vs GLM 5.2
Benchmarks through
Verdict
Gemini 3.1 Pro Preview places higher on 1 of the 1 tasks where both are ranked. GLM 5.2 costs 58.7x less per input token. Only GLM 5.2 has open weights for self-hosting.
- Choose Gemini 3.1 Pro Preview for Value LLM (#7).
- Choose GLM 5.2 for you need to self-host.
Built from the published task rankings below; ranks come from each task's own weighted benchmarks, so a win on one task says nothing about another.
Measured head to head
Benchmarks both models were run on
| Evaluation | Gemini 3.1 Pro Preview | GLM 5.2 | Source |
|---|---|---|---|
| Design Arena Coding | 1231 Elo | 1287 Elo | source · 2026-10-23 |
| Design Arena Full-stack Agent | 1068 Elo | 1241 Elo | source · 2026-10-22 |
| Design Arena UI | 1224 Elo | 1294 Elo | source · 2026-10-23 |
| Design Arena Web-app Agent | 1125 Elo | 1222 Elo | source · 2026-10-23 |
| Design Arena Website | 1234 Elo | 1285 Elo | source · 2026-10-23 |
| FrontierMath Tiers 1-3 (Epoch AI) | 59.6% ± 2.9 | 59.2% ± 3 | source · 2026-06-11 |
| GPQA Diamond (Epoch AI) | 94.4% ± 1.6 | 91.9% ± 1.6 | source · 2026-08-06 |
| GPQA Diamond (Epoch AI) | 94.1% ± 1.7 | 91.9% ± 1.6 | source · 2026-02-20 |
| LiveBench Agentic Coding | 44.1% | 51.8% | source · 2026-10-18 |
| LiveBench Coding | 76.5% | 79.7% | source · 2026-10-18 |
| LiveBench Data Analysis | 78.5% | 73.7% | source · 2026-10-18 |
| LiveBench Instruction Following | 79.1% | 62.3% | source · 2026-10-18 |
| LiveBench Language | 85.4% | 76.2% | source · 2026-10-18 |
| LiveBench Math | 91.0% | 89.8% | source · 2026-10-18 |
| LiveBench Reasoning | 84% | 78.6% | source · 2026-10-18 |
| LMArena Agent | -3.7 | 2.1 | source · 2026-10-20 |
| LMArena Agent Bash Recovery | -23.3 | 0.7 | source · 2026-10-08 |
| LMArena Agent Steerability | -5.5 | 1.3 | source · 2026-10-20 |
| LMArena Agent Task Outcome | -10.7 | 3.8 | source · 2026-10-20 |
| LMArena Agent Tool Hallucination | -0.1 | 0.7 | source · 2026-10-20 |
| LMArena Creative Writing | 1482 Elo | 1462 Elo | source · 2026-10-22 |
| LMArena Instruction Following | 1466 Elo | 1466 Elo | source · 2026-10-22 |
| LMArena Long Query | 1482 Elo | 1479 Elo | source · 2026-10-22 |
| LMArena Math | 1486 Elo | 1484 Elo | source · 2026-10-22 |
| LMArena Text | 1487 Elo | 1475 Elo | source · 2026-10-22 |
| LMArena Text Factuality | 1472 Elo | 1463 Elo | source · 2026-10-22 |
| LMArena WebDev | 1447 Elo | 1604 Elo | source · 2026-10-22 |
| OTIS Mock AIME (Epoch AI) | 95.6% ± 3.1 | 86.4% ± 4.2 | source · 2026-08-06 |
| OTIS Mock AIME (Epoch AI) | 95.6% ± 3.1 | 86.4% ± 4.2 | source · 2026-02-20 |
| SimpleQA Verified (Epoch AI) | 73.5% ± 1.4 | 34.2% ± 1.5 | source · 2026-08-10 |
| tau2-bench (1.0.1) | 26% | 37.1% | source · 2026-05-05 |
Values are as published by each evaluator, in its own units; rows from different evaluators are not on a common scale.
task evidence
How they place when the question changes
Lower ranks are better. “Best” marks the strongest option in this shortlist for that task.