SYSTEM ONE MODEL · BY TYPESAFE AI · 2026

Jev AI, decoded: pricing, access, benchmarks & news.

Jev doesn't write text — it returns typed decisions with calibrated probabilities. It's open to everyone now: grab $5 in free credit at console.typesafe.ai, or judge it yourself right here — no signup, 5 free runs a day. This hub tracks what it is, what it costs, and how to build on it — and keeps answering as things change.

193.6x
faster than frontier LLMs (self-reported)
444.6x
cheaper per workflow case (self-reported)
$0.042
per 1M input tokens — output free
$5
free credit — open access, no waitlist
01 · What is Jev

Jev, in 30 seconds

Jev reads a block of text (the state) plus questions you define, and returns one of three typed answers — a Choice from your options, an ordinal Score, or a Noul true/false probability — each with a confidence estimate. Because the answer shape is fixed in advance, it can't invent text, and it runs 20–200x faster than chat models at 40–400x lower cost. Your code sets confidence thresholds: act automatically when Jev is sure, escalate to a human or a frontier LLM when it isn't.

Read the full explainer →

02 · Latest news

Jev news, tracked daily

2026-11-18

So Was Jev Useful?

Editor’s take: Five points and a single comment is thin evidence for judging Jev's usefulness, so treat any verdict here as anecdote rather than a signal developers should act on. Read source →

Source: news.ycombinator.com
2026-11-11

Show HN: Thread Police eliminates stray replies in Slack channels with Jev

Editor’s take: A four-point, zero-comment HN post is barely a signal, so treat Jev's Slack threading demo as a promising pattern rather than proof the model handles messy real-world channel state. Read source →

Source: news.ycombinator.com
2026-11-11

Cutting LLM Token Costs with Currents News API and TypeSafe Jev Pre-Filtering

Editor’s take: Two points and zero comments is no validation at all, so treat the pre-filtering pattern as a design idea worth testing against your own Jev decision logs, not a proven cost saver. Read source →

Source: news.ycombinator.com

See the full timeline →

03 · Access

Start using Jev today

Jev opened to everyone on Sep 20 — no waitlist. Sign up directly, or go through a gateway:

04 · Pricing

Pricing at a glance

JevTypical chat LLM
Input$0.042 / 1M tokens$0.25 – $3.00 / 1M tokens
Output$0 (free)$2 – $15 / 1M tokens
Latency70–500 ms2 – 80 s
ReturnsTyped decision + confidenceFree-form text

Full provider comparison + cost calculator →

05 · Start here

Explore the hub

06 · FAQ

Jev AI FAQ — the short version

What is Jev in one sentence?
Jev is a "System One" AI model by TypeSafe that reads your input (state) plus typed questions and returns choices, scores, or true/false probabilities with confidence estimates — it never generates free-form text.
What are Jev's three question primitives?
Choice (pick one option from a list you define), Score (an ordinal rating such as low/medium/high/critical), and Noul (TypeSafe's binary type: the probability that a statement is true, e.g. 0.95). Multiple questions can share one request.
Do I have to wait on the TypeSafe waitlist?
No — as of Sep 20, 2026 the waitlist is gone. Jev is open to everyone: sign up at console.typesafe.ai and get $5 in free credit (~120M tokens). The Vercel AI Gateway (typesafe-ai/jev) and Cloudflare Workers AI (typesafe/jev) also serve it directly.
Which SDKs are available?
Official Python and JavaScript/TypeScript SDKs ship from TypeSafe, plus framework integrations: the Vercel AI SDK (via @ai-sdk/typesafe) and LangChain (langchain-typesafe package).
How does Jev's price compare to LLMs?
TypeSafe claims the $0.042/1M input rate is roughly 238x lower than "Claude Fable 5.1" and about 1/48th of GPT-5.6 Terra's $2.00/1M input — and those models also bill $2–$12 per 1M output tokens, which Jev doesn't charge at all. See our pricing page for a full comparison table.

All 20 questions, straight answers →