Up to four models
GLM-5.3-Flash vs Mistral Large 4
On the index (GPT-4 = 100), GLM-5.3-Flash scores 164 and Mistral Large 4 160. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GLM-5.3-Flash164148–180
WeightsOpen
ReleasedAug 2026
Price in / out$0.15 / $0.50
Context1.05M
Core tests taken6 of 9
Mistral Large 4160144–177
WeightsOpen
ReleasedOct 2026
Price in / out$1.36 / $4.18
Context1.05M
Core tests taken4 of 9
GLM-5.3-Flash and Mistral Large 4 are too close to call (4 points apart, within the uncertainty), ahead on 3 of 4 shared core tests, and costs 8.7× less.
Test by testthird-party results
GLM-5.3-FlashMistral Large 4Every result as a table
| Test | GLM-5.3-Flash | Mistral Large 4 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.1% | 88.1% |
| Maths | ||
| FrontierMath (tiers 1-3) | 55.8% | 58.6% |
| AIME-style maths (OTIS mock) | 93.9% | 98.3% |
| ProofBench v1.1 | 21.0% | 10.0% |
| Chess puzzles | 21.0% | 33.0% |
| FrontierMath (tier 4) | 17.1% | 26.8% |
| Surface Evolver Bench | 52.5% | 28.1% |
| Coding | ||
| Terminal-Bench 4.0 | 37.4% | 27.8% |
| Code Migration | 20.5% | 30.6% |
| IOI | 52.5% | 45.3% |
| Vibe Code Bench 1-100 | 16.0% | 14.3% |
| Vibe Code Bench v1.1 | 30.8% | 78.4% |
| Agents | ||
| Mystery game puzzles | 8.0% | 17.0% |
| Professional work | ||
| Finance Agent (v2) | 57.9% | 54.7% |
| EMB | 55.9% | 56.2% |
| HLAB-Vals-1.1 | 10.4% | 17.5% |
| Legal Research Bench | 45.2% | 31.7% |
| MedScribe | 88.9% | 80.4% |
| Tax Agent Bench | 31.6% | 30.4% |
Tests at least two of these models have taken. Bold rows are the core tests.
