ApprenticeBench: GPT-6 Astra learns on the job and beats human professionals
A new benchmark puts computer use and continual learning into a real job, claiming Astra and Fable 5.1 can learn on the job and surpass human practitioners.
Introducing ApprenticeBench: computer use + continual learning on a real job. We show Fable 5.1 and GPT-6 Astra can now continually learn on a job and surpass human professionals. A decisive step change in AI's job readiness. No FDEs. Agents deploy themselves into the job. 🧵
GUI vs. API, what's the future for computer-use agents (CUAs)? For the first time, we show quantitative evidence for the rapid progression of Claude and GPT models on CUA/GUI capabilities. Fable 5.1 and GPT-6 Astra can use GUIs just as well as APIs. No more CUA tax. Other models still have a large room to catch up. Grok 4.6 is strong with APIs but degrades by 71% on GUIs.
