Atlas

Benchmarks

← All benchmarks

APEX-Agents 1.1 Corporate Law (68-task image-free slice)

Professional Work

Mercor's APEX-Agents 1.1 Corporate Law job restricted to the 68 of its 80 tasks that carry no image inputs, a common slice Mercor defined so every model faces the same task set. Agents work in twelve worlds with files, mail, documents, spreadsheets, chat, calendar and code execution, and the score is the mean fraction of expert rubric criteria passed, averaged over four attempts per task before averaging tasks. Not comparable with the 80-task Corporate Law leaderboard or the 240-task APEX-Agents 1.1 headline.

Top models (higher is better)

ModelScore
GPT-6 Astra87.7
Claude Opus 5.585.4
GPT-6.1 Sol85.3
Claude Opus 583.8
Gemini 3.8 Flash82.7
Grok 4.682.2
GLM-5.381.4
GPT-5.6 Sol80.6
DeepSeek V4 Pro 081379.3
DeepSeek V4.1 Flash78.4
Kimi K376.8
Grok 4.775.6
Loading Atlas data…