OpenAI's Decisions API (public beta since Oct 6) and TypeSafe's Jev do the same job: closed questions in, typed answers with probabilities out, no prose.
Jev is cheaper ($0.042 vs $0.10 per million input tokens, no output charge on either) but text only. OpenAI takes images, offers US and EU residency, and can return a refusal. Jev always picks one of your options. Jev has a published calibration number; OpenAI has none yet.
My read: the price gap barely matters at agent volumes. What matters is what your code does when the model shouldn't be answering. A model that must choose gives you a confident label for input it didn't understand, so you need your own "insufficient" option and a human fallback either way.
I compared images, latency, pricing, residency, limits, and a few small tests that point in different directions here: OpenAI's Decisions API vs Jev: which one should you use?
Top comments (0)