Claude's Fable 5.2 beat OpenAI's GPT-6.2 Astra at real office tasks
What happened
Epoch AI graded six AI systems on tasks drawn from its own work. Claude Fable 5.2 beat GPT-6.2 Astra by 17 points, even though the two tie on a general benchmark. Astra averaged 65%.
Why it matters
Benchmark ties hide big gaps: Fable 5.2 scored 98% on data center analysis against Astra's 65%, so test models on your own tasks.
Impact
AI & developer tools · Software & SaaS
Benchmark ties hide gaps, so teams must test models on their own tasks.
Ask your team
Have we tested competing AI tools on our own actual work instead of trusting published benchmark scores?
Connecting the dots
- Before
- Epoch AI built a test from its own research work, graded by hand, where 100% equals the standard it expects of its staff.
- Next
- Watch for Epoch's full breakdown of scores across each task in its automation reports.
Tap an underlined word to see what it means.
Read the original at Epoch AI
Something wrong in this story? Report a mistake. We check it against the source within 24 hours.
Share with your take
Start from our line and make it yours. It stays on your device until you post it.
Ask about this story
Answered only from this story, with the line it came from.