Atlas

Benchmarks

← All benchmarks

Epoch Automation Reports — Research Design

Agents · 2026-10-08

Human rubric score on one research-design task with nine questions and human-intervention deductions.

Top models (higher is better)

ModelScore
Claude Fable 5.163.3
GPT-6 Astra61.7
Grok 4.655.0
Kimi K336.7
Gemini 3.8 Flash23.3
Loading Atlas data…