Up to four models
Gemini 3.5 Flash vs GLM-5.2
On the index (GPT-4 = 100), Gemini 3.5 Flash scores 162 and GLM-5.2 154. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
162146–177
WeightsClosed
ReleasedMay 2026
Price in / out$1.5 / $9.0
Context1.05M
Core tests taken7 of 9
154138–169
WeightsOpen
ReleasedJun 2026
Price in / out$0.0070 / $6.8
Context1.05M
Core tests taken6 of 9
Gemini 3.5 Flash leads GLM-5.2 by 8 points, ahead on 5 of 6 shared core tests, at 2.0× the price.
Test by testthird-party results
Every result as a table
| Test | Gemini 3.5 Flash | GLM-5.2 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 92.8% | 91.9% |
| GPQA Diamond (Vals AI) | 92.7% | 85.6% |
| MMLU Pro | 89.5% | 86.7% |
| EBR-bench | 4.8% | 9.5% |
| SimpleBench | 76.7% | 58.8% |
| SimpleQA Verified | 66.2% | 34.2% |
| Maths | ||
| FrontierMath (tiers 1-3) | 62.8% | 59.2% |
| AIME-style maths (OTIS mock) | 95.6% | 86.4% |
| Chess puzzles | 45.0% | 22.0% |
| FrontierMath (tier 4) | 26.8% | 29.3% |
| Surface Evolver Bench | 58.1% | 55.6% |
| Coding | ||
| SWE-bench Verified | 79.3% | 78.7% |
| Code Migration | 26.8% | 37.9% |
| LiveCodeBench | 87.6% | 69.5% |
| ProgramBench | 0.0% | 0.5% |
| SWE-bench | 78.8% | 82.8% |
| Vibe Code Bench v1.1 | 48.7% | 64.0% |
| DeepSWE | 37.4% | 43.8% |
| WeirdML | 62.6% | 70.1% |
| Agents | ||
| APEX-Agents | 27.5% | 45.2% |
| SkillsBench | 52.7% | 45.1% |
| DTBench | 94.7% | 93.6% |
| LMCA | 47.1% | 45.8% |
| Mystery game puzzles | 32.0% | 19.0% |
| Professional work | ||
| Finance Agent (v2) | 57.9% | 49.7% |
| CorpFin v2 | 64.7% | 66.1% |
| EMB | 63.5% | 61.5% |
| LegalBench | 83.6% | 84.1% |
| Legal Research Bench | 30.8% | 31.3% |
| MedCode | 55.8% | 40.8% |
| MedScribe | 76.6% | 83.5% |
| TaxEval v2 | 74.4% | 73.3% |
| Novel problems | ||
| ARC-AGI-2 | 72.1% | 22.8% |
| ARC-AGI-1 | 92.5% | 77.0% |
Tests at least two of these models have taken. Bold rows are the core tests.