Official benchmarks map / terminal-bench-science
moonshotai/Kimi-K3.1-Pro is ranked first on the terminal-bench-science leaderboard with 38.1, 195.3% ahead of deepseek-ai/DeepSeek-V4.1-Flash-1017. Read live from the Hugging Face API on November 6, 2026.
| # | Model | Organization | Score |
|---|---|---|---|
| 1 | moonshotai/Kimi-K3.1-Pro | moonshotai | 38.1 |
| 2 | deepseek-ai/DeepSeek-V4.1-Flash-1017 | deepseek-ai | 12.9 |
| 3 | zai-org/GLM-5.3 | zai-org | 9.5 |
| 4 | mistralai/Mistral-Large-4-1T-A52B | mistralai | 4.3 |
| 5 | inclusionAI/Ling-3.1-flash-1019 | inclusionAI | 4.3 |
| 6 | MiniMaxAI/MiniMax-M3.1-Flash | MiniMaxAI | 3.3 |
| 7 | stepfun-ai/Step-5-Preview | stepfun-ai | 2.4 |
| 8 | Qwen/Qwen3.8-27B | Qwen | 0.5 |
| 9 | reflection/Beam-1.0-501B-A23B | reflection | 0.5 |
terminal-bench-science is a dataset that Hugging Face marks as an official benchmark. Its leaderboard is built automatically from the .eval_results files that model repositories publish, so entries are mostly self-reported by model authors and evaluation settings differ between submissions. A verified tag appears only where Hugging Face marks an entry as verified.
Original board: huggingface.co/datasets/harborframework/terminal-bench-science
Part of the Hugging Face official benchmarks map by QuantID. Unofficial. Not affiliated with or endorsed by Hugging Face.