Equational Theories - Hard (Default reasoning): leaderboard
Metric: Strict F1 (%). Source: huggingface.co. Saturation forecast: Around January 2027. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 87.84 |
| 2 | Grok 4.1 Fast | 70.45 |
| 3 | Qwen 3.5 397B A17B | 65.51 |
| 4 | Kimi K2.5 | 65 |
| 5 | Qwen 3.5 122B A10B | 60.82 |
| 6 | Seed 2.0 Lite | 60.14 |
| 7 | Qwen 3.5 27B | 60 |
| 8 | GPT-5 Mini | 57.93 |
| 9 | GPT-OSS-120B | 56.73 |
| 10 | GLM-5 | 54.22 |
| 11 | Claude Sonnet 4.6 | 49.28 |
| 12 | Step 3.5 Flash | 46.43 |
| 13 | MiniMax-M2.5 | 46.27 |
| 14 | Llama 3.1 8B Instruct | 39.7 |
| 15 | Grok Code Fast 1 | 38.96 |
Interactive version: theaggregate.ai/benchmark?slug=equational-theories-hard-default-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-09.