ChatGPT Co-creator Introduces Jev: A New Type of AI Model for Instant Decisions
Diogo Almeida, one of ChatGPT's creators, unveiled Jev, a non-autoregressive AI model that outputs structured decisions instead of text, offering 20-200x speed and 40-400x cost reductions.
ChatGPT Co-creator Introduces Jev: A New Type of AI Model for Instant Decisions
Diogo Almeida, one of the creators of ChatGPT, has announced a fundamentally new type of AI model. After two years of secret development, Jev was revealed — a model that does not write text or generate tokens sequentially.
Instead, it takes any unstructured data and makes parallel decisions defined by the developer. For example, it selects one category from a list, assigns a score, or determines the probability of a statement.
Jev is also prohibited from using hidden chain-of-thought reasoning. Complex behavior must be assembled by the developer from many simple decisions using standard code.
This is why the model works 20–200 times faster and costs 40–400 times less. One million input tokens cost 4.2 cents, while the output remains free forever (its computation is so cheap that the company sees no point in charging for it).
The company claims zero hallucinations. Jev cannot produce an answer outside the specified options or violate the structure. However, the model can still choose the wrong option from the available ones.
This is not a replacement for GPT-6 Astra or a new chatbot. Jev is created for automation, where programs do not need text. They need fast decisions in a specified format.
Why it matters
AnalysisJev shifts the paradigm from generative text to parallel decision-making, offering a highly efficient solution for structured automation tasks where traditional LLMs are too slow and expensive.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Jev and System One AI: A Complete Practical Guide to the Decision Model from TypeSafe
A deep engineering breakdown of the flagship Jev model: System One architecture versus LLMs, Choice/Score/Noul primitives, request parallelism, confidence mathematics, video use cases, code, and an overview of local open-weight alternatives (Laya, GLiNER2.5, CLM-8B).
40+ Best Claude Skills for Creators, Brands, and Engineers
An encyclopedic catalog of over 40 practical Claude Skills: UI design, social media workflows, marketing, SEO, deep research, video and audio pipelines, product management, and engineering automation.
Jev Decision Model
Jev is a commercial non-autoregressive System 1 model from TypeSafe AI for fast typed decisions in agentic loops: it returns a validated decision object in 70–500 ms at ~$0.042/1M input tokens, eliminating JSON syntax hallucinations and high-latency routing in ReAct loop/Tool Calling stacks.
Tencent releases Hy4: an open-weight model challenging DeepSeek
Tencent has released Hy4, a multimodal model that scored 75.2 on APEX-Agents in internal tests, ahead of DeepSeek V4.1 Flash-1017. The company plans to publish the model weights.