LLMEval-Fair - Science: leaderboard
Metric: Discipline Score (0-10). Source: github.com. 59 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview) | 9.07 |
| 2 | DeepSeek R1 | 9.03 |
| 3 | Claude Sonnet 4.5 | 8.97 |
| 4 | DeepSeek V3 | 8.97 |
| 5 | Qwen 3 235B A22B | 8.97 |
| 6 | Claude Sonnet 4 (Thinking) | 8.93 |
| 7 | GPT-5 | 8.9 |
| 8 | O1 (2024-12-17) | 8.9 |
| 9 | Claude Sonnet 4.5 (Thinking) | 8.9 |
| 10 | GLM-4.6 | 8.9 |
| 11 | Gemini 2.5 Flash (Thinking) | 8.9 |
| 12 | DeepSeek V3.2 | 8.87 |
| 13 | Claude Sonnet 4 | 8.8 |
| 14 | Kimi K2 | 8.77 |
| 15 | Qwen 3 32B | 8.67 |
Interactive version: theaggregate.ai/benchmark?slug=llmeval-fair-science · How It Works · Data refreshed daily, snapshot 2026-09-19.