TypeSafe AI ships Jev, a model that returns typed probabilistic values instead of text, at $0.042 per million input tokens and free output
TypeSafe AI launched what it calls System One Models, an architecture that emits type-safe structured values with calibrated confidence scores rather than strings, and released Jev as the first public one on 2026-09-15. Founder Diogo Almeida frames it as "a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out," claiming 70-500ms end-to-end latency, parallel sampling of all outputs in a single query, and hallucination that is structurally impossible because the output type is guaranteed. Against GPT-6 Astra and Fable 5.1 on workflow evaluations the company reports roughly 193.6x faster and 444.6x cheaper; pricing is $0.042 per million input tokens with output free, behind an early-access waitlist. Treat the benchmark numbers as vendor-reported — but the shape is worth watching for anyone whose agent spends most of its tokens on routing and classification decisions rather than prose.
Source
↳ Follow the thread