Atlas

Benchmarks

← All benchmarks

Agents' Last Exam Unlicensed Overall Score

Agents · 2026-06-03

Average partial-credit rubric score across the 141-task unlicensed subset of Agents' Last Exam.

Top models (higher is better)

ModelScore
Claude Opus 5.565.9
GPT-6 Astra62.1
GPT-6 Sol61.7
Claude Opus 559.1
Muse Spark 1.358.4
MiniMax M3.1 Flash57.0
GPT-5.6 Sol56.0
GPT-5.6 Luna53.3
GPT-6 Luna53.1
Kimi K353.0
GPT-5.6 Terra52.5
Claude Fable 550.2
GPT-5.549.9
Opus 4.846.9
Agnes 2.5 Pro Beta46.1
Opus 4.744.7
GLM-5.242.9
Doubao Seed 2.1 Pro42.8
Composer 2.542.1
GPT-5.439.9
Gemini 3.1 Pro Preview34.8
Qwen3.7-Max (source unspecified)33.7
DeepSeek-V4-Pro29.7
GLM-5.129.7
Qwen3.6 Plus (2026-04-02)26.0
MiMo-V2.525.0
Kimi K2.622.9
Grok 4.322.8
MiniMax M2.714.9
Loading Atlas data…