Benchmark
SciCode
reasoning
math
physics
chemistry
code
text
SciCode is a research coding benchmark curated by scientists that challenges language models to code solutions for scientific problems. It contains 338 subproblems decomposed from 80 challenging main problems across 16 natural science sub-fields including mathematics, physics, chemistry, biology, and materials science. Problems require knowledge recall, reasoning, and code synthesis skills.
Top 10
榜单概览
数据说明
分数为 0-100 百分制综合成绩（Elo 榜为相对评分，仅供同榜内比较）。评测分数属事实数据，整理自公开评测体系与厂商公开资料，部分分数为厂商自报口径，不同榜单的评测方法与数据集可能存在差异，请结合多个榜单交叉参考。
Ranking
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | Fugu | Sakana AI | 60.1 | 来源 |
| 2 | Fugu Ultra | Sakana AI | 58.7 | 来源 |
| 3 | Qwen3.7 Max | Alibaba Cloud / Qwen Team | 53.5 | 来源 |
| 4 | GLM-4.5 | Zhipu AI | 41.7 | 来源 |
| 5 | Step 3.5 Flash | StepFun | 40.4 | 来源 |
| 6 | Step 3.7 Flash | StepFun | 40.0 | 来源 |
| 7 | Step 3.5 Flash 2603 | StepFun | 38.5 | 来源 |
| 8 | Qwen3 Max | Alibaba Cloud / Qwen Team | 38.3 | 来源 |
| 9 | Mistral Small 4 | Mistral AI | 38.0 | 来源 |
| 10 | GLM-4.5-Air | Zhipu AI | 37.3 | 来源 |
| 11 | Devstral 2 | Mistral AI | 33.1 | 来源 |