This browser is not keeping saved items, so they will be forgotten when you leave this page. Everything else still works.
Compare
Compare models and coding agents
Each ranking covers one benchmark. Models and complete agents or tools are compared separately, every figure links its source, and missing figures say so. How the rankings are made.
Model comparisons
Models measured on one benchmark at a time, through the tool the benchmark names.
- Aider polyglot (223 exercises)
1 results ranked by pass rate, second attempt, newest run .
- Aider polyglot (224 exercises)
5 results ranked by pass rate, second attempt, newest run .
- Aider polyglot (225 exercises)
63 results ranked by pass rate, second attempt, newest run .
Coding tool and agent comparisons
Complete agents and tools — their scaffold and model together — measured on one benchmark at a time.
- SWE-bench (Verified)
60 results ranked by resolved, newest run .