Start here / JevModel

What is JevModel, and how is it related to Jev?

JevModel is an independent Jev playground and API that calls the real typesafe/jev-1.13 model. Learn what it adds, what it costs, how data is handled and when to use the official TypeSafe console instead.

Last updated:

Short answer

  • JevModel is a website for using the Jev decision model without building anything first: a browser playground, a REST API, browser Batch and visual Decision Flows. It is independent and not affiliated with TypeSafe AI, which develops Jev.
  • Live answers come from the real typesafe/jev-1.13 model through OpenRouter. You can try 3 runs a day without signing up and get 10,000 free tokens after email verification.

What does JevModel provide?

JevModel is an independent hosted service for developers who want structured decisions from text or JSON. It provides a Playground, a REST API, saved visual decision flows and browser-based Batch processing. TypeSafe AI develops the underlying Jev model; JevModel is not affiliated with TypeSafe AI. This service currently calls typesafe/jev-1.13 through OpenRouter. It does not develop Jev, distribute its weights or offer model training.

What is the Jev model?Is Jev open source?Is Jev a Multimodal / Downloadable Model? What JevModel Actually Is

Why use JevModel instead of the official console?

TypeSafe AI runs its own playground and API in the TypeSafe console, and that is the right choice if you want first-party support and SDKs. JevModel exists for a different set of needs. You can see a real Jev answer on the homepage before creating any account. After email verification you get 10,000 free tokens. Non-developers can classify a CSV in the browser and export the answers. Teams can draw a multi-step Decision Flow and call it with one request. Token packs never expire, unused subscription tokens roll over, and Playground, Batch and API share one balance. These are additions on top of the same model, not a different model.

JevModel vs the official Jev playgroundHow to use Jev: a step-by-step guide

How do you know the answers are real?

Live runs are labelled as live in the playground and report the latency of the request. The server sends each validated request to OpenRouter with the pinned model ID typesafe/jev-1.13 and returns the model’s answers and token usage unchanged. Panels marked “illustrative” on the homepage are fixed examples, and the offline GitHub fixtures are illustrative too; neither is presented as a model result. JevModel does not publish its own accuracy benchmark, so measure quality on your own labeled cases.

Evaluate a Jev workflowJev probability, confidence and human-review thresholds

What goes in, and what comes back?

Send one state and one to eight typed questions to POST /api/v1/systemone. A state is text, a JSON object or a nonempty JSON array. Choice selects from your named labels, Score evaluates an ordered rubric and can be fractional, and Noul reports a yes probability. The API wraps answers in { code, message, data }; data includes request_id, answers, usage and quota. Your application interprets the answer and performs any action.

Choice, Score, and NoulJevModel API: request and responseHow to classify text with the JevModel API

How do Playground, Flows and Batch differ?

Use Playground to inspect a request before integration. A Flow saves questions and branches; only visited questions execute and the host application acts on the returned outcome. Batch reads CSV, TXT or JSONL in the browser, sends row requests and exports answers. It is not a durable asynchronous bulk-job API, and closing the browser can interrupt processing. Console and API share the same account allowance and token balance.

AI Decision Flows with Jev and human reviewBatch text classification with the JevModel APIJevModel pricing and free Jev runs

What does JevModel cost?

The homepage demo allows 3 runs per UTC day without an account. Verified accounts get 10,000 free tokens once, shared across Playground, Batch, Flows and API. Each successful request deducts the input tokens the model reports from your balance; output is free and failed requests are refunded. Token packs start at $4.90 and never expire; monthly plans keep unused tokens, and yearly plans are 50% off with every token delivered up front. Any purchase also unlocks unlimited Decision Flows and 1,000-row batches. These are JevModel prices and are separate from TypeSafe’s direct pricing.

JevModel pricing and free Jev runs

Who runs JevModel, and how is data handled?

JevModel is operated independently and has no partnership with TypeSafe AI or OpenRouter beyond being an API customer. Inputs pass through this site and OpenRouter to the model. No-signup runs are not stored. Signed-in runs are kept in your history for up to 30 days, limited to the latest 100, and you can delete them at any time. The service depends on OpenRouter continuing to offer typesafe/jev-1.13. Do not send secrets or sensitive personal data.

Private runs and history

When should you use something else?

Use a text generator when the product needs a written answer, a multimodal model when it needs image or audio evidence, and deterministic rules for exact permissions or financial constraints. This service does not provide local inference, fine-tuning, a first-party Python or JavaScript SDK, or a privacy guarantee for data sent upstream. Its public example clients use ordinary HTTP and are independent of the official TypeSafe SDK. Check the data policy before submitting sensitive inputs.

Jev vs GPT and Claude: choosing a decision layerJev or rules?Private runs and history

How can you check it before adopting it?

Start with the offline GitHub examples, which require no key or model request. Sign in, verify your email and create a JevModel key for live API testing. Evaluate representative labeled inputs, ambiguous cases and failures before enabling automation. Public examples contain illustrative answers; this site does not publish an independently measured accuracy, latency or cost advantage over alternatives.

Your first Jev decisionEvaluate a Jev workflow
git clone https://github.com/SouthCat196/jevmodel-examples.git
cd jevmodel-examples
node quickstart/node.mjs

Frequently asked questions

Is JevModel made by TypeSafe AI?

No. TypeSafe AI develops the Jev model. JevModel is an independent service that calls typesafe/jev-1.13 through OpenRouter.

Is JevModel free?

You can run Jev 3 times a day without signup, and verified accounts get 10,000 free tokens. Further runs use tokens from a pack or plan.

Does JevModel use the real Jev model?

Yes. Live runs call typesafe/jev-1.13 and return its answers unchanged. Only panels marked illustrative show fixed examples.

Can I use JevModel without writing code?

Yes. The playground, Batch and Decision Flows editor work in the browser. Code is only needed to call the API.

JevModel is independent and not affiliated with TypeSafe AI.