Benchmarks
Coding benchmarks
These benchmarks compare models on programming tasks. Some score the code itself. Others ask people to choose the better answer.
We track 299 benchmarks in this topic. Main benchmarks are listed by how many models they cover, most first. Where a benchmark has separate results for a subset of its tasks, those results appear right under it. Click one for its leaderboard.
Showing 101–200 of 299 benchmarks.
- Claw-Eval24 models
- AlgoTune18 models
- CodeForces17 models
- LiveCodeBench AtCoder16 models
- LiveCodeBench Codeforces16 models
- LiveCodeBench Easy16 models
- LiveCodeBench LeetCode16 models
- LiveCodeBench Medium16 models
- KernelBench Pass Rate15 models
- KernelBench Peak Performance15 models
- CAD-Eval14 models
- hal-swe-bench-verified-mini12 models
- ExploitGym11 models
- hal-scicode11 models
- Aider-Polyglot Edit10 models
- BenchCAD (with Python tool)8 models
- Ramp SWE-Bench v2 (Bash)7 models
- Ramp SWE-Bench v26 models
- SWE-Lancer (IC-Diamond subset)6 models
- FrontierCode 2.05 models
- Aider4 models
- FrontierCode4 models