# TypeSafe AI — Jev

> AI lab announcements, news reports and podcasts.

forck.live follows what companies announce about their own products. Press may
corroborate an official story; it never leads and never appears as an item of its
own. A missing date stays missing. A title is the publisher's words — where a
company does not publish in English, an English rendering is listed separately
as `Title in English` and named as this site's translation, never as theirs.

This file is the record of one model version: what the model registry states
about it, TypeSafe AI's launch post, the first-party posts this record
holds about it, the outlets that reported on those posts, and other companies'
announcements that it runs on their platforms. A model ranks nothing here and
nothing is composed: each date names whose it is, and a field nothing states is
`not stated` rather than missing.

- This model's page: https://forck.live/models/jev
- TypeSafe AI's page: https://forck.live/company/typesafe-ai
- Every model: https://forck.live/models
- Catalogue: https://forck.live/llms.txt

## Model

- Name: Jev
- Company: TypeSafe AI (typesafe.ai)
- Launch post: Introducing System One Models & Jev — https://forck.live/items/12212-introducing-system-one-models-jev
- Announced: 2026-09-15
- Launch source: https://typesafe.ai/blog/introducing-system-one-models-and-jev
- Launch chosen by: a person, pinned in the model registry
- Released (per models.dev): not stated
- Listed (per OpenRouter): not stated
- Open weights: not stated
- Posts about it: 3
- Outlets named under its posts: 4
- Also written: jev-latest; jev-preview
- Variant: Jev Router — https://forck.live/models/jev-router
- Variant: Jev 1.13 — https://forck.live/models/jev-1-13
- Price: Input tokens: $0.042 / MTok ($42 per billion tokens). Output tokens: FREE (too cheap to meter). — from https://forck.live/items/12212-introducing-system-one-models-jev

## Posts about it

_TypeSafe AI's own posts whose title names Jev, and its launch post, in the record's order: dated newest first, undated last. The record reads the newest 50 first-party posts mentioning the model, every company's together, so a model mentioned more often than that lists the newest of them._

Every entry is one first-party announcement, and carries the same eight fields
in this order before any others:

    Company · Announced · Category · Coverage · Announcement · Group · Source · Record

`Source` is the company's own page the announcement was traced to and `Record`
is this site's page for it. Both are safe to cite.

`Coverage` is a number of outlets, or `not counted` — which means this record
reached no count of its own, never that nobody wrote about it.

`Category` is this tracker's own analysis, and is authoritative wherever it is
stated. `not stated` means either that nothing has analysed the post yet or
that the analysis placed it outside this taxonomy; neither is a verdict about
the announcement.

`Announcement` is whether the post is news of something the company did, rather
than a how-to, a customer story, a changelog listing or a staffing note.

`Group` is which of this record's groups the entry falls in, carried on the
entry itself so it holds where the groups are headings and where they are not.
One of: `models`, `covered`, `announcements`, `routine`.

Anything after `Record` is held detail that only some announcements carry.

### TypeSafe AI — Introducing System One Models & Jev

- Company: TypeSafe AI (typesafe.ai)
- Announced: 2026-09-15
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://typesafe.ai/blog/introducing-system-one-models-and-jev
- Record: https://forck.live/items/12212-introducing-system-one-models-jev
- Subject: Jev / System One

TypeSafe AI released its first System One Model, Jev, in early access. Jev is optimized for structured outputs and calibrated probabilistic decisions, claims to be two orders of magnitude faster and more efficient than existing LLMs on System One tasks, and cannot hallucinate or make type errors.

### TypeSafe AI — Jev: For the builders AND dreamers

- Company: TypeSafe AI (typesafe.ai)
- Announced: not stated by the source (first seen 2026-10-14T23:05:25.851211+00:00)
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://typesafe.ai/blog/jev-1-14
- Record: https://forck.live/items/22458-jev-for-the-builders-and-dreamers
- Subject: Jev / System One

TypeSafe AI released Jev 1.14.0 with small improvements across the board and introduced experimental features under a "danger zone" flag for users who want to push boundaries. The company says it balances reliability for production builders with creativity for dreamers, offering image support, a new bounding box type, and longer context as previews. Image support works with the existing Jev 1.13.0 model, which the company says was multimodal all along, and is composable via find-and-replace in assets. The new box type returns a single bounding box and is available only for Jev 1.14.0, while longer context is now opt-in despite the company's recommendation to limit it for reliability. Jev 1.14.0 is available under jev-preview, with jev-latest updating in a week, and Jev 1.13.0 will remain available with 60 days notice before deprecation.

### TypeSafe AI — How Jack & Jill Cut Candidate Screening Costs By 88% With Jev

- Company: TypeSafe AI (typesafe.ai)
- Announced: not stated by the source (first seen 2026-10-14T22:02:24.783063+00:00)
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://typesafe.ai/blog/jack-jill-jev-case-study
- Record: https://forck.live/items/22433-how-jack-jill-cut-candidate-screening-costs-by-88-with-jev
- Subject: Jev / System One

TypeSafe AI published a case study showing that talent marketplace Jack & Jill replaced Google's Gemini 3.1 Flash Lite with TypeSafe's Jev model for candidate screening. Jev maintained similar accuracy while reducing costs by 88% and cutting median screening time by roughly 50%.

## Who reported on it

_Outlets this record can name and link, under the first-party post each one covered. Their number is `Outlets named under its posts` above, and the rows are not totalled again. A post's own `Coverage` can be higher: it counts outlets the signal saw that this record cannot name and link._

### On: Introducing System One Models & Jev

Record: https://forck.live/items/12212-introducing-system-one-models-jev

- RuntimeWire — TypeSafe opens Jev early access for fast, typed AI decisions: https://runtimewire.com/article/typesafe-jev-system-one-ai-model-early-access
- Seangoedecke — Jev means structured output is interesting again: https://seangoedecke.com/jev-means-structured-output-is-interesting-again
- The Neuron — TypeSafe JEV Explained: AI Decisions Without a Chatbot: https://theneuron.ai/explainer-articles/typesafe-jev-system-one-models-explained
- The Rundown AI — TypeSafe launches Jev for AI decisions inside software: https://therundown.ai/news/typesafe-jev-ai-decisions-software

## Where it runs

_Other companies' own announcements naming Jev and TypeSafe AI — each is that company's post, not TypeSafe AI's._

### Cloudflare — Introducing Clairic, a new System One class of extraction models

- Company: Cloudflare (cloudflare.com)
- Announced: 2026-10-19T17:48:26+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://blog.cloudflare.com/clairic-extraction-model/
- Record: https://forck.live/items/24693-introducing-clairic-a-new-system-one-class-of-extraction-models
- Subject: Cloudflare AI

Cloudflare is previewing Clairic, the first of a new class of System One extraction models, hosted on Workers AI as @cf/cloudflare/clairic. Clairic extracts, computes, estimates and forecasts values from unstructured inputs, which Cloudflare says is more accurate, faster and cheaper than traditional LLMs for those tasks. It follows TypeSafe's Jev and Clef, which Cloudflare describes as decision models that return probabilities against listed questions; Clairic instead answers with typed values. An extraction model returns extracted values such as names or order IDs as spans with character offsets, computed numbers, dates and lists, estimated values as a distribution with a median and an 80% interval, and forecasted values one step ahead. A request is a state of text plus up to four optional images and a map of typed fields, each with a name, type and optional instructions, answered in one parallel pass. Because it generates no tokens, Cloudflare says it cannot mistype an answer or hallucinate the way an LLM can, though it can still make mistakes, and it returns null when no answer is found. Cloudflare says it intends to release the model open-weights soon, and that Clairic charges only for input tokens and typically responds in under a second.

### Shanghai AI Lab — 看见即行动、性能超Jev，上海AI实验室推出决策模型“书生·明决”，实现响应效率与质量双升级

- Company: Shanghai AI Lab (shlab.org.cn)
- Announced: 2026-10-12
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.shlab.org.cn/news/5444306
- Record: https://forck.live/items/19747-jev-ai
- Title in English (translated by forck.live from zh): See and act, performance surpasses Jev: Shanghai AI Lab launches decision model 'Shusheng Mingjue', achieving dual upgrades in response efficiency and quality
- Subject: InternLM / 书生

Shanghai AI Lab released Intern-Decision (Shusheng Mingjue), a decision-making model that outperforms Jev and other open-source models on seven benchmarks, with three parameter sizes: 0.8B, 2B and 4B. The model integrates native visual perception with instruction following, enabling it to make decisions directly from visual input without separate image processing. On a single RTX 4090, the 0.8B and 2B models achieve about 30 decisions per second with roughly 33 ms latency, while the 4B model averages 44.16 ms — more than twice as fast as Jev's 109.70 ms. The 4B model scored 90.02 average across seven tasks, 1.28 points higher than Jev's 88.74, and topped the leaderboard. The team also explored a confidence-based routing mechanism where Intern-Decision handles routine judgments and passes uncertain cases to a larger model, cutting total task time by about 53% while losing only two correct answers.

### Vercel — Jev for Python engineers

- Company: Vercel (vercel.com)
- Announced: 2026-10-02T07:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/jev-for-python-engineers
- Record: https://forck.live/items/20649-jev-for-python-engineers
- Subject: v0 / AI SDK / AI Gateway

Vercel released an update to its AI SDK for Python, adding an experimental evaluate() API that interfaces with the Jev classifier model. The API allows Python developers to use Jev for narrow decision-making tasks by feeding it data and multiple-choice questions, receiving structured JSON responses with confidence levels.

### Vercel — Jev is the fastest-adopted model in AI Gateway history

- Company: Vercel (vercel.com)
- Announced: 2026-09-18T07:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/ai-gateway-jev-model-launch
- Record: https://forck.live/items/20654-jev-is-the-fastest-adopted-model-in-ai-gateway-history
- Subject: v0 / AI SDK / AI Gateway

Vercel reports that Jev, a probabilistic decision model from TypeSafe AI, became the fastest-adopted model on AI Gateway within 24 hours of launch, reaching more than twice as many paid teams as any previous model. By hour 24, nearly 13% of paid teams were using it, outpacing GPT-5.6 and Fable 5.1. TypeSafe AI states Jev is up to 194 times faster and 445 times cheaper than language models in its own evaluations.

---

Record: https://forck.live/models/jev
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md
