TypeSafe AI

TypeSafe AI provides a decision API built around its Jev model. An application can send a support message and ask which department should handle it. The application defines the possible answers; Jev returns a choice and probabilities. Open the Jev model page for a walkthrough, interactive examples and model-specific limits.

What TypeSafe’s API returns

The official API accepts a state and a map of typed questions through POST /v1/systemone. Choice selects among specified options, Score evaluates a rubric, and Noul returns a yes/no probability. The answers are intended for application code to consume directly. These primitives can be combined in one request.

Published price and limits

The current model documentation lists jev-1.13.0 at $0.042 per million input tokens, with output tokens free. Checked October 8, 2026. This rate is known; catalog-only status does not mean the provider has an unknown price.

The documented context budget is 64,000 tokens for state and all questions together, with a separate 32,000-token limit for state plus the longest question. Those two limits cannot be accurately represented by the calculator's single chat prompt and response profile.

Why it is not in calculated stacks

The current workload form describes generated chat responses and research language calls. It does not collect the size of a shared state, the number and size of typed questions, or the longest question. Applying its usual chat token assumptions would obscure the actual workload and suggest that Jev generates a chat answer.

A calculated entry needs a decision workload that preserves these inputs and checks both context limits. Until then, use the API documentation and measured input-token usage to plan this part of an application. There is no calculator lock button or inferred task-quality score. Provider claims about speed and accuracy are not substituted for independent evidence in the ranking.