MathConstraint: leaderboard
Metric: Accuracy (self-reported). Source: benchmarklist.com. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 66.9 |
| 2 | Grok 4.20 | 58.1 |
| 3 | Gemini 3.1 Pro (Preview) | 51.7 |
| 4 | Qwen 3.6 Plus | 47.4 |
| 5 | Claude Opus 4.7 | 42.6 |
| 6 | GPT-OSS-120B | 39.5 |
| 7 | DeepSeek V4 Pro | 39.5 |
| 8 | Kimi K2.6 | 35.9 |
| 9 | DeepSeek V4 Flash | 34 |
| 10 | Gemini 3.1 Flash Lite | 24.9 |
| 11 | Claude Sonnet 4.6 | 18.5 |
| 12 | Llama 3.3 70B Instruct | 8.5 |
Interactive version: theaggregate.ai/benchmark?slug=mathconstraint · How It Works · Data refreshed daily, snapshot 2026-09-05.