Benchmark
Aider-Polyglot
general
code
text
A coding benchmark that evaluates LLMs on 225 challenging Exercism programming exercises across C++, Go, Java, JavaScript, Python, and Rust. Models receive two attempts to solve each problem, with test error feedback provided after the first attempt if it fails. The benchmark measures both initial problem-solving ability and capacity to edit code based on error feedback, providing an end-to-end evaluation of code generation and editing capabilities across multiple programming languages.
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 | GPT-5 | OpenAI | 88.0 | 来源 |
| 2 | DeepSeek-V3.2-Exp | DeepSeek | 74.5 | 来源 |
| 3 | DeepSeek Reasoner | DeepSeek | 74.2 | 来源 |
| 4 | DeepSeek Chat | DeepSeek | 70.2 | 来源 |
| 5 | Kimi K2-Instruct-0905 | Moonshot AI | 60.0 | 来源 |
| 6 | Qwen3-Next-80B-A3B-Instruct | Alibaba Cloud / Qwen Team | 49.8 | 来源 |