vibedonaldsvibedonalds.com
Jev logo

Jev

TypeSafe's System One model: an API that answers typed questions with probabilities instead of generated text.

What is Jev, in two sentences

Jev is TypeSafe's first public System One model, an API that takes a state and typed questions and returns choices, scores or 0–1 truth values with probabilities, not generated text. It costs $0.042 per million input tokens ($42 per billion), and output tokens are free.

About Jev

TypeSafe calls System One models a new class of AI model built for decisions inside software. Each request sends a state, as a string or a structured object, with any mix of Choice, Score and Noul questions: a Choice picks an option from a list, a Score rates the state on a rubric you define, and a Noul gives a 0–1 value for a true/false statement. Choice and Score answers also include probabilities and a confidence value, and every question is evaluated in parallel against the same state.

Jev is not a chat or code-completion model: TypeSafe's docs say it does not generate text, write code or hold a conversation, and it is not a drop-in replacement for the LLM behind Claude Code, Cursor or similar tools. The intended pattern is to have your coding agent write code that calls Jev, and TypeSafe offers an agent skill for Claude Code, Codex and other agents, Python and JavaScript/TypeScript SDKs, an HTTP endpoint and a Playground that runs Jev in the browser.

The current models are Jev 1.14 (model ID jev-1.14.0, available from October 14, 2026) and Jev 1.13, at the same price, each with a context of 64k tokens per request and 32k per question; a long-context request header raises both to 262k tokens. Rate limits are set per organization, with higher limits on custom and enterprise plans. Jev reads text sent as a string, a JSON object or an array, and the models page says it does not accept audio or video.

With Jev 1.14, TypeSafe also opened experimental features behind 'danger zone' API headers: images in requests, a Box question type that returns a single bounding box, and longer context, which it says it cannot yet vouch for in production. The vendor says Jev is not trained on customer requests or responses and that the same weights serve every account. Jev was introduced on September 15, 2026, and TypeSafe's homepage now says it is available to everyone.

Sources: docs.typesafe.ai/introduction, typesafe.ai, /introduction/quickstart, /models, /introduction/coding-agents, typesafe.ai/blog/introducing-system-one-models-and-jev and typesafe.ai/blog/jev-1-14, read 2026-10-27

Jev pros and cons

Pros

  • TypeSafe guarantees that answers fit the schema you define: given a list of categories, Jev cannot invent one outside it.
  • Many questions can share one request, and TypeSafe says adding questions barely changes the response time.
  • Probabilities and confidence let your code set thresholds for when to act on its own and when to ask for review.
  • You adapt Jev to a domain through the request, with rules in each question's instructions and criteria and your records in the state.

Cons

  • It cannot write text or code, so it cannot be the model behind a chatbot or a coding agent.
  • TypeSafe says tasks that need extended reasoning, such as complex mathematics or chess-like planning, may suit large reasoning models better.
  • The shape of an answer is guaranteed, its correctness is not: Jev can still choose the wrong option.
  • Accuracy is best in English; TypeSafe says other languages are handled but not equally well.

Tagged

  • Agentic Mode

Jev alternatives

The closest alternatives to Jev.

  1. 01
    Cloudflare Clairic

    Cloudflare's AI model for fast, cheap data extraction

    →
  2. 02
    Claude Haiku 5.5

    Anthropic's fastest and most capable Haiku yet

    →
  3. 03
    Mellum by JetBrains

    Fast LLMs for low-latency and high-performance workflows

    →

Related reading

Reviews
No reviews yet — be the first to rate Jev.

Featured on Vibedonalds

Own Jev? Add this badge to your site to show you’re listed — and link back to your profile here.

Jev — Featured on Vibedonalds
<a href="https://vibedonalds.com/tools/jev" target="_blank" rel="noopener nofollow">
  <img src="https://vibedonalds.com/badge/featured-on-vibedonalds.svg" alt="Jev — Featured on Vibedonalds" width="240" height="60" loading="lazy" />
</a>

Frequently asked questions

What is Jev?
Jev — TypeSafe's System One model: an API that answers typed questions with probabilities instead of generated text.
Is Jev free?
Jev is a paid product. Pricing details are on the official site.
What are the alternatives to Jev?
The closest alternatives to Jev are Cloudflare Clairic, Claude Haiku 5.5, Mellum by JetBrains.
What are the downsides of Jev?
It cannot write text or code, so it cannot be the model behind a chatbot or a coding agent. TypeSafe says tasks that need extended reasoning, such as complex mathematics or chess-like planning, may suit large reasoning models better. The shape of an answer is guaranteed, its correctness is not: Jev can still choose the wrong option. Accuracy is best in English; TypeSafe says other languages are handled but not equally well.
Is Jev an LLM?
TypeSafe says Jev is neither small nor an LLM: it understands language but does not generate free-form text or work as a chatbot. It answers the structured questions you send with typed decisions, probabilities and confidence.
How much does Jev cost?
Jev is billed per input token at $0.042 per million, or $42 per billion, and output tokens are free; Jev 1.14 and Jev 1.13 are priced the same. Higher rate limits are available on custom and enterprise plans through TypeSafe's sales team.
Is Jev open source or on GitHub?
TypeSafe serves Jev through its own API, and its models page says the same weights serve every account but mentions no weights to download. What TypeSafe points to on GitHub is its agent skill for coding agents, installed with npx skills add typesafe-ai/skills.
Can Jev replace the model in my coding agent?
TypeSafe's docs say Jev is not a drop-in replacement for the LLM behind Claude Code, Cursor and similar tools, and no model setting turns a coding agent into a Jev-powered one. Your coding agent can instead write code that calls Jev, with TypeSafe's agent skill giving it context on the API.
Has TypeSafe published benchmarks for Jev?
TypeSafe says it deliberately chose not to publish results on public benchmarks. It points to its own workflow evals instead and notes that they were made by people on its model capabilities team, so some bias could exist.