September 19, 2026
AI apps get typed decisions: Jev reached 13% of teams on its first day
According to Vercel, on September 16, 2026, Jev reached about 13% of AI Gateway teams on its first day after launch. The model returns typed Choice, Score, and Boolean values with probabilities instead of generating strings. In AI Gateway, it costs $0.04 per 1 million input tokens. TypeSafe lists $0.042, free output, and 70–500 ms latency.
Guillermo Rauch
@rauchg
The data and individual Jev adoption stories are staggering. Everyone is adopting it. I think it is a great product, but it is also a consequence of the sentiment that AI is too expensive and slow. People are ready to optimize processes and embed AI in even more places!
Jev was adopted faster than any other model in AI Gateway history. In the first day, @typesafeai reached ~13% of teams, 2x the GPT-5.6 family and 6x Fable 5.1.
· 132.4K views
Previously, these scenarios used LLMs for text-based evaluations inside workflows. Jev now returns a typed decision: it selects the next tool or subagent, determines whether to continue an operation, assesses urgency or risk, and verifies a model response.
Integration. AI Gateway requires a key from the Vercel Dashboard in the `AI_GATEWAY_API_KEY` variable, which the AI SDK picks up automatically. Then install `ai@latest` with `pnpm add ai@latest` and call `experimental_evaluate` with the `typesafe-ai/jev` model, passing `state` and `questions`. The `evaluate` API is supported starting with AI SDK 7.0.105.
TypeSafe launched Jev in early access on September 15, 2026. According to TypeSafe data cited by Vercel, Jev was up to 193.6 times faster and up to 444.6 times cheaper than LLMs for workflow evaluations.
As of September 16, 2026, Jev remains in early access.
Source
