Open CoT - LSAT Analytical Reasoning: leaderboard
The "LSAT Analytical Reasoning" subtask of Open CoT. A task from the Open CoT Leaderboard, measuring chain-of-thought reasoning ability and the accuracy gain from step-by-step reasoning.
Metric: CoT Gain (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 132 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 Distill Llama 70B | 22.61 | #594 |
| 2 | DeepSeek R1 Distill Qwen 14B | 16.96 | #844 |
| 3 | DeepSeek R1 Distill Qwen 32B | 16.52 | #655 |
| 4 | Llama-3.1-SuperNova-Lite | 13.91 | #730 |
| 5 | Llama 3.1 8B Instruct | 11.74 | #1035 |
| 6 | Qwen 2.5 14B Instruct | 11.74 | #647 |
| 7 | merlinite-7B | 10.43 | #1232 |
| 8 | Qwen 2.5 7B Instruct | 9.13 | #862 |
| 9 | Mixtral 8x7B Instruct (v0.1) | 8.7 | #1073 |
| 10 | Llama 3.1 Nemotron 70B Instruct | 8.7 | #554 |
| 11 | SuperNova-Medius | 8.7 | #479 |
| 12 | Llama 3 8B Instruct | 8.26 | #1129 |
| 13 | DeepSeek R1 Distill Llama 8B | 8.26 | #1291 |
| 14 | Llama 3 70B | 8.26 | #631 |
| 15 | internlm2-chat-20B | 8.26 | #999 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-cot-lsat-analytical-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.