Benchmarks

SWE-bench Multilingual

Codingunit % · normalized over [0, 92]

SWE-bench Multilingual — 300 real-issue problems spanning 9 programming languages, extending SWE-bench beyond Python.

#ModelSourceScoreNormalized

Score distribution

8 tracked results across the normalization window

092

Score vs. parameters

Open-weights models, log-x params