Atlas

Benchmarks

← All benchmarks

Agents' Last Exam Unlicensed Near-Term Pass Rate

Agents · 2026-06-03

Perfect-score pass rate on the 67-task unlicensed near-term subset of Agents' Last Exam.

Top models (higher is better)

ModelScore
Claude Opus 5.553.7
GPT-6 Astra52.2
GPT-6 Sol50.7
Claude Opus 549.3
GPT-5.6 Luna49.3
GPT-5.6 Sol47.8
Muse Spark 1.347.8
MiniMax M3.1 Flash46.3
GPT-5.6 Terra44.8
GPT-5.544.0
Opus 4.843.3
Kimi K340.3
GPT-6 Luna38.8
Claude Fable 537.3
Composer 2.534.3
GPT-5.433.6
GLM-5.232.8
Agnes 2.5 Pro Beta31.3
Opus 4.731.3
Gemini 3.1 Pro Preview28.4
Doubao Seed 2.1 Pro26.9
GLM-5.120.1
DeepSeek-V4-Pro19.9
Qwen3.7-Max (source unspecified)17.9
Kimi K2.615.7
Qwen3.6 Plus (2026-04-02)12.7
MiMo-V2.511.9
Grok 4.310.4
MiniMax M2.710.4
Loading Atlas data…