World signalVerified
OpenAI GDPval: the best models approach expert quality on tasks from 44 jobs
On one-off tasks written by professionals, like a legal brief or an engineering blueprint, top models come close to experts, so juniors will be judged on what they add beyond a first draft.
From Campus Binge's data, as of 2025-09. Each fact says how far to trust it.
When
2025-09
VerifiedOpenAI, GDPval
Where
us
VerifiedOpenAI, GDPval
Direction
headwind
VerifiedOpenAI, GDPval
Weight
matters
VerifiedOpenAI, GDPval
Sectors
AI and work, Law
VerifiedOpenAI, GDPval
Careers
2Sources
- OpenAI, GDPval Verified
Block signal/openai-gdpval-2025 · sha256:e195927318b0… · JSON · Markdown