Atlas

Benchmarks

← All benchmarks

SWE-Marathon - release-reported version unspecified

Code

Model-card quoted SWE-Marathon score without identified version or trial artifact. Diagnostic when the current v1.1 source does not reproduce that claim; native effort is preserved independently.

Top models (higher is better)

ModelScore
Muse Wonder Full57.5
Muse Wonder53.1
Kimi K3.1 Pro51.7
Loading Atlas data…