Epoch tests Fable 5.2 and GPT-6.2 Astra on workplace tasks

epoch.ai17 min read

Researchers evaluated several frontier AI models on updated workplace tasks, finding that none met human employee standards across the board. While models like Fable 5.2 and Opus 5.5 reached contractor-level performance in AI data center analysis, they consistently struggled with open-ended research design, effective iteration, and audience alignment. Open article →