New to this? Read it in simple words
- On September 15, TypeSafe AI launched Jev, a new kind of AI model. It gives choices and probabilities instead of writing text.
- Jev costs $0.042 per million input tokens, and output is free. Its launch post says large language models charge $0.20 to $10 per million input tokens.
- Vercel says teams started using Jev faster than any other model on its AI Gateway. But Jev was free on Vercel at the time.
- It is not yet clear if the price covers the real cost of running Jev. TypeSafe’s speed and cost figures have not been checked by outside groups.
- Token
- A small piece of text, often part of a word. AI use is priced per token.
- Probability
- A number that shows how likely something is.
- Large language model
- An AI model trained on huge amounts of text to read and write language.
TypeSafe AI came out of stealth on September 15 with a new kind of AI model called Jev. It does not write sentences. Instead, it answers questions from other software and gives a probability for each answer. The company calls it a “System One” model, a name taken from psychologist Daniel Kahneman’s idea of fast, intuitive thinking.
Former OpenAI researchers built it for software, not people
TypeSafe was founded in 2024 and is based in San Francisco. Its chief executive, Diogo Almeida, is a former OpenAI researcher. He is a co-author of the InstructGPT paper, which helped lead to ChatGPT.
The other co-founders are Erik Gafni, the chief technology officer, and Sasha Sheng, the chief operating officer. The company says Jev is built for software to use directly, not for people to chat with.
DCVC led a $40 million seed round. Forbes reported a valuation of $200 million, citing a person familiar with the deal. TypeSafe has not named the other investors in the round.
TypeSafe charges $0.042 per million input tokens. It says it cannot yet prove the price is not subsidised.
The price is tiny, and answers are free
TypeSafe charges $0.042 per million input tokens, or $42 per billion. A token is a small piece of text, often part of a word. Output is free, which the company calls “too cheap to meter”.
The launch post compares this with large language models that charge $0.20 to $10 per million input tokens. It adds that their output tokens usually cost about five times more than input.
The company admits one risk in plain words: “We can’t prove it isn’t subsidized.” Only time will show whether the price covers the real cost of running the model.
Access is still limited. TypeSafe is moving developers from a waitlist into early access. Jev is also listed on Vercel’s AI Gateway, OpenRouter, and Cloudflare Workers AI, and it is free on Vercel until September 25.
Fast adoption, measured during a free week
Vercel says Jev became the fastest-adopted model in the history of its AI Gateway. By hour 24, nearly 13% of paid teams were using it.
That was twice the share of OpenAI’s GPT-5.6 family, and more than six times the share of Anthropic’s Claude Fable 5.1.
The numbers need context. Jev was free on Vercel at the time, so trying it cost teams nothing. SiliconANGLE also notes that the company’s speed and cost figures have not been independently verified.
The next tests are simple. Watch for general access, the price after the free period, and more results from outside groups.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2
- 3
- 4
- 5
- 6Research · September 16, 2026TypeSafe AI exits stealth with $40M to build AI for use by software SiliconANGLE
Price, speed and usage figures come from the companies. The valuation comes from a source quoted by Forbes.
Former OpenAI researchers launched Jev, a model that decides. Silicon AI News, September 23, 2026. https://siliconainews.com/stories/typesafe-jev-launch
TypeSafe launched Jev with a $40 million seed round and very fast early adoption, though that measure came during a free week. How we rate the mood
The mood shows how the news frames the event for the tech and AI world: as an opportunity (6 and above) or as a loss (below 5). It is not our verdict on who is right.