Independent guide · Jev / TypeSafe

Jev developer guide (unofficial)

Q: What is the latest Jev model ID?

It is jev-1.13.0 on TypeSafe's own API, and the aliases jev-latest and jev-preview both point to it (checked 25 Sep 2026). Other channels use their own IDs: typesafe/jev-1.13 on OpenRouter, typesafe-ai/jev on Vercel AI Gateway and typesafe/jev on Cloudflare Workers AI.

By Ben (BenX) · Last checked against official sources 25 Sep 2026

This independent guide covers Jev, the decision model from TypeSafe AI: what it is, the current model ID, what it costs, where you can call it, and which of our guides answers your question.

Short answer: Jev is TypeSafe AI's decision model: you send text (the state) plus typed questions, and it returns typed answers with probabilities instead of generated text. The latest model ID is jev-1.13.0, and the jev-latest alias points to it. You can call it directly at POST https://api.typesafe.ai/v1/systemone, or through OpenRouter, Vercel AI Gateway or Cloudflare Workers AI.

Independent guide. aiedu.guide is not affiliated with TypeSafe AI or Jev, and is not endorsed or sponsored by either. Product names belong to their owners, and the official sources we checked are listed at the end.

Jev at a glance

Item	Jev 1.13 (last checked 25 Sep 2026)
Latest model ID	`jev-1.13.0` on TypeSafe; the aliases `jev-latest` and `jev-preview` both point to it. On other channels: `typesafe/jev-1.13` (OpenRouter), `typesafe-ai/jev` (Vercel AI Gateway), `typesafe/jev` (Cloudflare Workers AI).
Launch date	Announced 15 Sep 2026, in early access
Price per 1M tokens (input / output)	$0.042 / free, per TypeSafe. $0.042 per 1M input tokens on Vercel and Cloudflare, checked 28 Sep 2026.
Context	64k tokens per request in total, and 32k for the state plus the longest question, per TypeSafe. OpenRouter and Cloudflare list 32,000 tokens.
Rate limits (source)	TypeSafe's Models page: 250,000 tokens per second and 1,200 requests per minute, which TypeSafe says can change without notice. We found no Jev-specific limit on the gateway pages we checked.
Where it runs	TypeSafe directly, OpenRouter, Vercel AI Gateway and Cloudflare Workers AI
Input	Text only
Last checked	25 Sep 2026

Figures are TypeSafe's and each platform's own listings, not our measurements. Details for each channel are in Jev model IDs by channel.

Which guide do you need?

How do I make my first Jev API call?
The endpoint, a first curl request, the Python and TypeScript SDKs, and how to read the answer.
How do I get a Jev API key, and which route should I use?
The four routes compared: TypeSafe direct, OpenRouter, Vercel AI Gateway and Cloudflare Workers AI.
Should I use Jev or an LLM?
When a decision model beats a text model, when it's the wrong tool, and how to use both.
Which question type should I use: Noul, Choice or Score?
What each type returns, when to use it, and the mistakes that break thresholds.
What is the model ID and field name on OpenRouter, Vercel or Cloudflare?
Model IDs, endpoints, noul vs boolean, where confidence lives, and version pinning on each channel.
What confidence threshold should I use?
What confidence means, act and review bands for each question type, and why TypeSafe sets no default.
Why am I getting a 401, 422, 429 or 529 error?
What each Jev API error code means and what to do next.
What does Jev cost, and what are the limits?
TypeSafe's listed price, rate limits and context length.

What changed recently?

These are dated changes to Jev and its channels, taken from the official pages listed below. Times are as published.

21 Sep 2026: Vercel AI Gateway added a TypeSafe-client-compatible API and an HTTP API (POST https://ai-gateway.vercel.sh/v1/evaluate) for Jev. The Python SDK typesafe-sdk 0.7.1 was released on PyPI (UTC date).
18 Sep 2026: OpenRouter's typesafe/jev-1.13 model page lists this as the release date.
17 Sep 2026: TypeSafe last reviewed its jev-1.13 jaggedness page, which lists the model's known weak spots.
16 Sep 2026: Vercel announced Jev on AI Gateway as typesafe-ai/jev.
15 Sep 2026: TypeSafe announced Jev in early access. The JavaScript SDK @typesafe-ai/sdk 0.6.0 was published on npm (UTC date).

FAQ

What is Jev?

Jev is a decision model from TypeSafe AI, the first of its "System One" models. It returns typed answers with probabilities (a yes/no probability, a choice from options you define, or a score on levels you define) instead of generating text.

What is the latest Jev model ID?

It is jev-1.13.0 on TypeSafe's own API, and the aliases jev-latest and jev-preview both point to it (checked 25 Sep 2026). Other channels use their own IDs: typesafe/jev-1.13 on OpenRouter, typesafe-ai/jev on Vercel AI Gateway and typesafe/jev on Cloudflare Workers AI.

How much does Jev cost?

TypeSafe lists Jev 1.13 at $0.042 per million input tokens ($42 per billion), and output tokens are free. $0.042 per 1M input tokens on Vercel and Cloudflare, checked 28 Sep 2026. Prices can change, so check each platform's own page before you budget.

What are Jev's rate limits?

TypeSafe lists 250,000 tokens per second and 1,200 requests per minute, and says these limits can change without notice. A request over either limit returns a 429 error.

Where can I call Jev?

On four documented channels: TypeSafe directly, OpenRouter, Vercel AI Gateway and Cloudflare Workers AI. Only TypeSafe issues a TypeSafe key; the other three use their own keys and billing.

Is aiedu.guide the official Jev site?

No. aiedu.guide is an independent, unofficial guide. It is not affiliated with, endorsed by or sponsored by TypeSafe AI.

Official sources checked

Facts on this page come from these pages, all checked on 25 Sep 2026. Jev launched in September 2026 and details change quickly, so check the official page before you rely on a number.