Atlas

Benchmarks

← All benchmarks

Code Migration v1.1 - Code Quality

Code · 2026-10-19

Code-quality rating (0-5) on Code Migration v1.1 CLI tasks: the mean of four reviewer dimensions (readability, documentation, file structure and language idioms), scored independently of the hidden-test pass rate. Split out because it is a 0-5 rubric, not a 0-100 pass rate.

Top models (higher is better)

ModelScore
Claude Haiku 5.54.1
Claude Sonnet 5.53.8
GLM-5.33.8
DeepSeek V4.1 Flash3.7
MiMo-V2.6-Pro3.6
Step 5 Preview3.6
MiMo-V2.6-Flash3.6
GLM-5.3 Flash3.6
DeepSeek V4.1 Flash 10173.5
GPT-6.1 Sol3.4
Hy4 Preview3.4
GPT-6 Astra3.4
Kimi K33.3
DeepSeek V4 Pro 08133.3
Muse Wonder Full3.2
Grok 4.73.2
Mercury 33.2
MiniMax M33.1
Gemini 4 Argon3.0
Muse Wonder3.0
Gemini 3.8 Flash2.9
GPT-6 Luna2.9
Muse Spark Open2.4
Beam 1.0 501B A23B2.1
Inkling-Small2.0
Inkling2.0
Loading Atlas data…