Equational Theories Benchmark: leaderboard
Metric: F1 (%, mean of six settings). Source: huggingface.co. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 87.02 |
| 2 | Qwen 3.5 122B A10B | 76.03 |
| 3 | Qwen 3.5 397B A17B | 75.74 |
| 4 | GPT-5 Mini | 70.12 |
| 5 | Qwen 3.5 27B | 70.11 |
| 6 | GPT-OSS-120B | 68.59 |
| 7 | Seed 2.0 Lite | 64.5 |
| 8 | Claude Sonnet 4.6 | 60.95 |
| 9 | Claude Opus 4.6 | 60.7 |
| 10 | Gemini 3 Flash (Preview) | 60.22 |
| 11 | Kimi K2.5 | 57.24 |
| 12 | GLM-5 | 52.77 |
| 13 | GPT-5.4 | 52.12 |
| 14 | Grok Code Fast 1 | 47.34 |
| 15 | Grok 4.1 Fast | 46.88 |
Interactive version: theaggregate.ai/benchmark?slug=equational-theories-benchmark · How It Works · Data refreshed daily, snapshot 2026-09-05.