An independent analysis suggests Jev is not the frontier model it was thought to be, with only 4 to 9 billion active parameters
Jev is a decision model launched in September by TypeSafe AI, founded by former OpenAI researcher Diogo Almeida: it does not generate text but returns probabilistic answers from predefined options. The company pitched it as a “frontier-intelligence function call.” JevResearch, which says it is unaffiliated with TypeSafe, ran 16,379 live requests and concluded that the model most plausibly has 4 to 9 billion active parameters. Jev scored 21.9% on Humanity’s Last Exam, versus 49.5 to 62.5% for frontier models. The estimate relies only on API behavior, and a larger MoE design is not ruled out. TypeSafe has not disclosed a parameter count.
Sources
- TypeSafe AI, “Introducing System One Models & Jev”, (typesafe.ai)
- JevResearch (GitHub), “Jev: Not Frontier, But Still Worth Your Attention”, (jevresearch.github.io)
- JevResearch (GitHub), “Jev-Research”, data and code repository for the report, (github.com)
About this story
This story was posted on Instagram by @jarrus.tech on Oct. 5, 2026.
Spotted an error in this story? [email protected] · Instagram
Short link: thejarrus.com/en/h11-17
This story in Turkish: Bağımsız bir analiz Jev’in sanıldığı gibi sınır seviyesinde bir model olmadığını, yalnızca 4 ila 9 milyar aktif parametreye sahip olduğunu öne sürdü
Related
TypeSafe, the maker of Jev, is in talks at 50 times last week’s valuation
TypeSafe, the maker of Jev, raised a $40 million seed at a $200 million valuation on September 15 and is now discussing a valuation above $10 billion.