Equational Theories - Hard (Low-or-none reasoning): leaderboard
Metric: Strict F1 (%). Source: huggingface.co. Saturation forecast: Not forecast. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 122B A10B | 59.14 |
| 2 | Gemini 3.1 Pro (Preview) | 58.61 |
| 3 | GPT-OSS-120B | 56.97 |
| 4 | Qwen 3.5 27B | 56.68 |
| 5 | Qwen 3.5 397B A17B | 56.1 |
| 6 | Claude Opus 4.6 | 53.9 |
| 7 | GPT-5 Mini | 51.91 |
| 8 | Claude Sonnet 4.6 | 50 |
| 9 | Grok Code Fast 1 | 45.07 |
| 10 | Gemini 3 Flash (Preview) | 40.65 |
| 11 | Seed 2.0 Lite | 38.71 |
| 12 | Llama 3.1 8B Instruct | 33.47 |
| 13 | Claude Haiku 4.5 | 27.48 |
| 14 | GPT-5 Nano | 26.93 |
| 15 | MiniMax-M2.5 | 19.75 |
Interactive version: theaggregate.ai/benchmark?slug=equational-theories-hard-low-or-none-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-09.