Jev

TypeSafe AI · proprietary · jev

Jev is TypeSafe AI's first "System One Model": a non-autoregressive frontier model that takes unstructured program state as input and outputs type-safe, calibrated structured decisions (with confidence scores) in a single parallel pass instead of generating text token-by-token. It is trained with a method TypeSafe calls Reinforcement Learning for Calibrated Decisions (RLCD). TypeSafe reports end-to-end response times of 70ms-500ms (vs. 3-329s for frontier LLMs on comparable tasks) and says schema conformance is guaranteed by construction (no type errors). Launch pricing is $0.042 per million input tokens with output tokens unmetered ("free"). Available today in early access via TypeSafe's console/API; no downloadable weights, published parameter count, or independently verified benchmark scores were available at launch.