Epoch Automation Reports — Data Insights
Agents · 2026-10-08
Mean human rubric score on three data-insight generation tasks, with task-specific section weights and human-intervention deductions.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4.6 | 78.6 |
| GPT-6 Astra | 77.4 |
| Claude Fable 5.1 | 69.8 |
| Kimi K3 | 61.4 |
| Gemini 3.8 Flash | 58.3 |