Equational Theories - Hard2 (Low-or-none reasoning): leaderboard
Metric: Strict F1 (%). Source: huggingface.co. Saturation forecast: Not forecast. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 91.19 |
| 2 | Qwen 3.5 397B A17B | 71.52 |
| 3 | Qwen 3.5 122B A10B | 70.13 |
| 4 | Qwen 3.5 27B | 69.17 |
| 5 | GPT-OSS-120B | 64.88 |
| 6 | GPT-5 Mini | 64.38 |
| 7 | Claude Opus 4.6 | 59.71 |
| 8 | Llama 3.1 8B Instruct | 59.55 |
| 9 | Seed 2.0 Lite | 58.17 |
| 10 | Claude Sonnet 4.6 | 54.82 |
| 11 | Grok Code Fast 1 | 50.23 |
| 12 | GPT-5 Nano | 35.35 |
| 13 | Gemini 3 Flash (Preview) | 30.53 |
| 14 | MiniMax-M2.5 | 24.63 |
| 15 | Claude Haiku 4.5 | 18.68 |
Interactive version: theaggregate.ai/benchmark?slug=equational-theories-hard2-low-or-none-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-09.