TypeSafe AI has opened public access to its Jev model, following the removal of a waitlist just five days after its initial launch on September 15, 2026. The company, founded by Diogo Almeida, Gafni, and Sheng, previously secured $40 million in seed funding led by DCVC.
Jev is designed for rapid, structured decision-making rather than general text generation. It processes inputs with latencies between 70 and 500 milliseconds and supports a context window of approximately 32,000 tokens. New users receive $5 in credits, with input costs set at $0.042 per million tokens.
The model has already integrated with infrastructure providers including LiteLLM, Netlify AI Gateway, Cloudflare, OpenRouter, and Vercel. While performance in classification tasks is noted, independent testing has identified vulnerabilities to adversarial inputs that may lead to inaccurate decision outcomes.
There are currently no documented production-level benchmarks or regulatory reviews for the model. The startup continues to address questions regarding the deployment of its System One architecture within automated agent systems.
