The Jev model from TypeSafe AI is now available for free in beta via the Venice application programming interface. This specialized tool returns typed answers, such as a probability or a rubric score, instead of generating the conversational text typical of large language models. Users can implement the model for tasks like routing support tickets or identifying fraudulent signups through a dedicated decisions endpoint.

Deployments of Jev on the Vercel AI Gateway and OpenRouter are priced at $0.042 per 1 million input tokens with free output. OpenRouter reports end to end latency between 70 and 500 milliseconds and claims the tool is up to 190 times faster and 440 times cheaper than using standard LLM logic. The model supports a 64K token context window per call.