Jev vs Jev Router
OpenRouter lists Jev Router under author Typesafe. This page is unofficial and implies no partnership with TypeSafe or OpenRouter.
Jev is TypeSafe’s hosted System One model. Jev Router is a separate OpenRouter listing (typesafe/jev-router) that OpenRouter says runs on Jev.
This unofficial note restates what OpenRouter and TypeSafe publish. It does not rank products or infer routing rules that those pages do not state.
What each listing is
Jev is TypeSafe’s flagship System One model. TypeSafe’s docs describe a hosted call: send a state plus typed questions, receive structured answers. Official usage is POST https://api.typesafe.ai/v1/systemone (or a TypeSafe SDK) with model jev-latest. TypeSafe currently documents text input only (strings, JSON objects, and arrays of text).
Jev Router is an OpenRouter model listing. The official model page and the OpenRouter models API both identify it as typesafe/jev-router, display name TypeSafe: Jev Router, author Typesafe. OpenRouter’s copy:
Jev Router picks the best model and reasoning effort for each request, balancing quality, speed, and cost. It runs on Jev, TypeSafe’s first System One model, and adapts as your conversation evolves. One endpoint gives you the whole model ecosystem, leveraging the ecosystem’s token and spend share.
OpenRouter presents it as an OpenAI-compatible chat completion: you set the model slug to typesafe/jev-router. That is a different interface from TypeSafe’s POST /v1/systemone decision API.
Relationship (only what official text says)
OpenRouter’s model page and FAQ state that Jev Router runs on Jev, and that Jev is TypeSafe’s first System One model. The same FAQ says Jev 1.13 is another text model from Typesafe. OpenRouter’s Typesafe author page lists Jev Router, Jev Latest, and Jev 1.13 as separate Typesafe models.
TypeSafe’s own docs describe Jev and a pattern of using Jev to build a custom router (classify intent, estimate difficulty, escalate). They do not, in the public TypeSafe documentation checked for this page, publish a product page for the OpenRouter slug typesafe/jev-router. Beyond OpenRouter’s “runs on Jev” sentence, the two vendors do not jointly spell out ownership, billing of the routed model, or whether the listing is a TypeSafe-operated service.
This site therefore treats them as related listings, not as the same API.
What OpenRouter publishes for Jev Router
Queried from openrouter.ai/typesafe/jev-router and GET https://openrouter.ai/api/v1/models (search typesafe/jev-router).
| Field | Official value |
|---|---|
| Model id / canonical slug | typesafe/jev-router |
| Display name | TypeSafe: Jev Router |
| Author / provider on OpenRouter | Typesafe (author page) |
| Context length | 1,000,000 tokens (page FAQ and context_length in the models API) |
| Pricing on the model page | FAQ: pricing shown is zero; you are not charged for prompt or completion tokens. Page JSON-LD offers.price is 0 USD. |
| Pricing in the models API | pricing.prompt and pricing.completion are the string "-1" |
Inputs / output (page FAQ + API architecture) | Audio, files (page: such as PDFs), images, text, video → text |
| Released (page FAQ) | September 25, 2026 |
API created | Unix 1790363560 |
| Tokenizer (models API) | Router |
| Endpoints listing | GET /api/v1/models/typesafe/jev-router/endpoints returned "endpoints": [] |
At the same query, the public models catalogue contained one id matching typesafe / jev: typesafe/jev-router. Sibling names on the author page were not present as separate rows in that catalogue snapshot.
How routing is described
Official model page (primary source): picks the best model and reasoning effort for each request; balances quality, speed, and cost; adapts as the conversation evolves; one endpoint for the model ecosystem.
The page does not list eligible models, weights for those tradeoffs, cache keys, developer routing controls, or fallbacks. This site does not fill those gaps.
Announcement (secondary): OpenRouter’s post x.com/OpenRouter/status/2103610898690855161 introduces typesafe/jev-router as “a cache-aware model router powered by Jev” and repeats the quality / speed / cost sentence. That “cache-aware” wording is not on the official model page. The page does not say the router selects a model from cache state.
Surfaces that differ in the docs
| TypeSafe Jev (TypeSafe docs) | Jev Router (OpenRouter listing) | |
|---|---|---|
| Product kind | Hosted System One decision model | OpenRouter model slug that routes to a model and reasoning effort |
| Call shape | POST /v1/systemone with state + typed questions | OpenRouter OpenAI-compatible chat API, model typesafe/jev-router |
| Documented output | Structured answers (choice / score / noul, probabilities) | Text (page FAQ) |
| Documented input | Text | Audio, files, images, text, video |
| Context (as published) | Not the 1,000,000-token figure; OpenRouter’s Jev 1.13 page lists 32,000 tokens | 1,000,000 tokens on the Jev Router listing |
A 1,000,000-token window on the router listing is not explained as applying to Jev itself, to the routed model, or to both.