Up to four models
GLM-5.3-Flash vs Gemini 3.6 Flash
On the index (GPT-4 = 100), GLM-5.3-Flash scores 161 and Gemini 3.6 Flash 161. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
161145–177
WeightsOpen
ReleasedAug 2026
Price in / out$0.15 / $0.50
Context1.05M
Core tests taken6 of 9
161145–176
WeightsClosed
ReleasedJul 2026
Price in / out$0.75 / $3.75
Context1.05M
Core tests taken5 of 9
GLM-5.3-Flash and Gemini 3.6 Flash are too close to call (1 point apart, within the uncertainty), ahead on 3 of 5 shared core tests, and costs 6.3× less.
Test by testthird-party results
Every result as a table
| Test | GLM-5.3-Flash | Gemini 3.6 Flash |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.1% | 94.1% |
| GPQA Diamond (Vals AI) | 86.4% | 93.4% |
| MMLU Pro | 86.1% | 89.3% |
| MMMU Pro | 86.0% | 88.4% |
| Maths | ||
| FrontierMath (tiers 1-3) | 55.8% | 59.0% |
| AIME-style maths (OTIS mock) | 93.9% | 94.2% |
| Chess puzzles | 21.0% | 42.0% |
| FrontierMath (tier 4) | 17.1% | 21.9% |
| Coding | ||
| Code Migration | 20.5% | 30.9% |
| CyberBench v1.1 | 67.9% | 45.4% |
| IOI | 52.5% | 35.1% |
| LiveCodeBench | 80.5% | 88.1% |
| ProgramBench | 0.0% | 0.0% |
| SWE-bench | 92.0% | 79.6% |
| Vibe Code Bench v1.1 | 30.8% | 64.0% |
| DeepSWE | 63.4% | 46.7% |
| Agents | ||
| APEX-Agents | 52.8% | 46.9% |
| Terminal-Bench-Science | 2.9% | 4.3% |
| Mystery game puzzles | 8.0% | 30.0% |
| Professional work | ||
| Finance Agent (v2) | 57.9% | 56.3% |
| EMB | 55.9% | 65.4% |
| LegalBench | 83.9% | 86.7% |
| Legal Research Bench | 45.2% | 25.0% |
| MedScribe | 88.9% | 79.7% |
| Tax Agent Bench | 31.6% | 18.0% |
| TaxEval v2 | 75.6% | 74.9% |
| Novel problems | ||
| ARC-AGI-2 | 65.8% | 60.4% |
| ARC-AGI-1 | 91.0% | 91.2% |
Tests at least two of these models have taken. Bold rows are the core tests.