Science Leaderboard: leaderboard
Science Leaderboard (TIGER-Lab) — multi-domain science QA aggregate; average accuracy.
Metric: Average Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 Turbo | 67 | #404 |
| 2 | Llama 3 70B Instruct | 61.9 | #634 |
| 3 | Llama 3 8B Instruct | 45.4 | #1129 |
| 4 | internlm2-20B | 43.3 | #943 |
| 5 | Mixtral 8x7B Instruct | 42.6 | #1170 |
| 6 | Yi 34B (Base) | 39.9 | #972 |
| 7 | deepseek-llm-67B-base | 39.4 | #973 |
| 8 | internlm2-7B | 31.2 | #1192 |
| 9 | Yi 6B (Base) | 23.6 | #1472 |
| 10 | Llama 2 13B Base | 22.9 | #1445 |
| 11 | deepseek-llm-7B-base | 21.7 | #1443 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=science-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.