Epoch AI tests workplace automation with real-job tasks
Epoch AI expanded its workplace-automation experiment into the Epoch Automation Reports, evaluating models on recurring work such as graphic design and research-project planning against the standards expected of human employees. Claude Fable 5.1 and GPT-6 Astra led with an average score of 65%, but Epoch AI said both models struggled with judgment-heavy work, actionable experiments, and polished graphic design. Epoch AI has partly automated basic chart generation because Fable 5.1 usually produces an acceptable chart on its first attempt.
The evaluation provides real-work evidence that strong benchmark scores do not necessarily translate into tasks that can be handed over to AI without human review.
This source provides an excerpt. Continue reading on the original site.
#Epoch AI#Epoch Automation Reports#Claude Fable 5.1#GPT-6 Astra