Reaction
Epoch says Fable 5.2 tops its work test—but still falls below the employee standard
Epoch reports Fable 5.2 led its work-task benchmark at 82%, while GPT-6.2 Astra placed fourth.
TLDR
Fable 5.2 led Epoch’s work-task benchmark at 82%, while GPT-6.2 Astra placed fourth at 65%, Epoch reports. An employee grades the outputs against internal quality standards; 100% represents the expected employee standard. No tested model reached it. Fable and Astra share the top ECI score of 174, but that tie didn’t carry over to the work tasks.
Combined views
10
5 Sources, first seen ago

Comments