Atlas

Benchmarks

← All benchmarks

Code Migration - CLI

Code · 2026-06-16

The CLI split of Code Migration. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.

Top models (higher is better)

ModelScore
Claude Sonnet 5.569.8
Gemini 4 Argon67.6
GPT-6 Astra67.5
GPT-6.1 Sol63.5
Muse Wonder60.1
Muse Wonder Full58.0
Claude Opus 553.3
GPT-6 Sol52.9
GPT-5.6 Sol47.2
GPT-5.6 Terra40.4
Muse Spark 1.340.4
Opus 4.840.1
Hy4 Preview39.9
Grok 4.639.4
DeepSeek V4.1 Flash37.5
GPT-6 Luna37.2
GPT-5.536.9
Grok 4.736.4
Sonnet 536.3
Claude Haiku 5.536.1
GPT-5.6 Luna36.1
GLM-5.335.6
DeepSeek V4.1 Flash 101735.6
Opus 4.735.2
MiMo-V2.6-Pro34.0
DeepSeek V4 Pro 081332.1
DeepSeek-V4-Flash-073132.0
Step 5 Preview32.0
MiMo-V2.6-Flash31.2
Sonnet 4.630.3
GLM-5.227.2
Gemini 3.8 Flash25.9
Grok 4.525.5
GPT-5.423.8
Gemini 3.7 Flash23.5
Ember-123.3
Gemini 3.6 Flash19.3
GLM-5.118.8
Muse Spark 1.118.6
Kimi K2.618.0
Loading Atlas data…