Decision model Jev breaks adoption records amid hype and skepticism
Jev, the first public model from TypeSafe AI, has become the hottest topic in AI circles recently: a "decision-only" model that does no text generation. It takes a state (webpage, ad, ticket, DOM, etc.) and a typed question, and returns only a choice from a predefined option set with a calibrated confidence score—it is physically incapable of outputting anything off-schema. Per an AGI Hunt teardown, Jev claims to be 20-200x faster and up to 400x cheaper than generative LLMs, priced at $0.042 per million tokens. Official Vercel data shows about 13% of teams adopted it on day one on the AI Gateway—the fastest adoption of any model in the gateway's history, at 2x the pace of the GPT-5.6 series over the same period and 6x that of Fable 5.1. Within 36 hours of launch, 140,000 people joined the waitlist, and it held the HN top spot for a full day.
Confirmed
- Vercel officially announced Jev set the AI Gateway adoption-speed record: 13% of teams onboarded on day one—2x the GPT-5.6 series and 6x Fable 5.1.
- Jev is temporarily free on the Vercel AI Gateway, with promotional pricing through September 25, 2026; the company positions it as a "System 2"-style judgment primitive.
- Company background: founded by Diogo Almeida, a former OpenAI researcher and co-creator of RLHF (some posts also call him a "co-creator of ChatGPT"); seed round of $40 million led by DCVC. DCVC investor James Hardiman said the company's vision 18 months ago was "intelligence primitives reliable, fast, and cheap enough to be everywhere."
- Community momentum: 140,000 waitlist signups in 36 hours, a full day at #1 on HN, and 359 community projects already built; someone wrote a 34-page hands-on guide with 20 workflow designs, viewing it as a possible component of the 2028 AI engineer stack.
Views and Controversy
- @signulll analyzed that Jev-style models have no decoding loop, bounded outputs, and parallelizable inference, making them ideal for on-device use—embedding "judgment" into all kinds of devices at zero marginal cost and near-zero latency, and thinks this "will happen soon."
- @dpopa relayed a long-form take: traditional software achieves responsiveness via 50-100ms event loops, while autoregressive models take 1-5 seconds to emit text; Jev's millisecond-scale decision engine could reshape generative GUIs and end text-prompt-dominant interaction.
- @Loran750 threw cold water: what Jev does could have been done years ago with fixed embeddings plus logistic regression—same input, same output, thresholds under your control; the hype just made everyone "discover" the category of classifiers. Others claimed it renders half of AI engineering obsolete.
Why it matters
- Jev represents the breakout of a new category: "judgment/decision primitives." Structured output, calibrated confidence, and extremely low latency and cost could reshape the cost structure of the many classification and routing scenarios in LLM applications; but its true increment over traditional classifiers—and the claims of 200x faster and 400x cheaper—still await independent verification.
2026-09-19 ~ 2026-09-21 · 14 related posts
- Episode 1: TypeSafe exits stealth with decision model Jev and RLCD training method(2026-09-16, 86 posts)
- Episode 2: TypeSafe AI launches Jev, a dedicated evaluation model showing major speed and cost gains in tests(2026-09-16, 8 posts)
- Episode 3: Vercel fx to adopt Jev safety reviewer, up to 18x faster(2026-09-17, 3 posts)
- Episode 4: OpenJev Open-Source Clone Runs Jev-Style API on One RTX 3090(2026-09-17, 3 posts)
- Episode 5: TypeSafe AI Launches Jev, a Decision-Only Model That Outputs No Text(2026-09-17, 73 posts)
- Episode 6: Mystery decision model Jev sparks wave of community benchmarks(2026-09-18, 6 posts)
- Episode 7: TypeSafe AI's Jev model and RLCD method spark debate(2026-09-18, 3 posts)
- Episode 8: LangChain Publishes Tutorial on Building an Agent Harness with Jev(2026-09-18, 7 posts)
- Episode 9: Jev Ecosystem Explodes with Wave of Popular GitHub Projects(2026-09-18, 4 posts)
- Episode 10: ChatGPT co-inventor launches Jev, a decision-only AI model(2026-09-18, 34 posts)
- Episode 11: TypeSafe Jev Decision Models Spark Debate on Breakthrough, Skepticism and Replication(2026-09-18, 7 posts)
- Episode 12: Jev Matches Classifiers at 100x Lower Cost(2026-09-19, 2 posts)
- Episode 13: Jev Positioned as LLM Router, Not a Replacement(2026-09-19, 2 posts)
- Episode 14: Developers Hail Jev as Top Value Model for Batch Labeling(2026-09-19, 2 posts)
- Episode 15: JevBench Launches: Jev 1.13 Holds Narrow Lead as Open-Source Rivals Close In(2026-09-19, 8 posts)
- Episode 16: Decision model Jev breaks adoption records amid hype and skepticism(2026-09-19, 14 posts)
- Episode 17: Researcher Tests Jev as a New Primitive: Cheap Verifiers Enable Self-Checking Long-Horizon Agents(2026-09-19, 10 posts)
- Episode 18: 16-model calibration test: Jev fastest honest model but ranks 10th in accuracy(2026-09-19, 8 posts)
- Episode 19: Independent Tests Confirm Jev Classifier's Accuracy at Fractional Cost(2026-09-19, 4 posts)
Primary sources
- Why Jev-class models could become a near-free judgment primitive running on-device — signulll · 2026-09-19
- Jev hits fastest model adoption in AI Gateway history, reaching ~13% of teams on day one — HankYeomans · 2026-09-19
- Jev becomes fastest-adopted model in Vercel AI Gateway history, used by 13% of paid teams in 24 hours — multiply_matrix · 2026-09-19
- [source] Eval model Jev goes free on Vercel AI Gateway until Sept 25 — cramforce · 2026-09-19
- TypeSafe's first model Jev sets fastest adoption record in AI Gateway history within 72 hours — hackgoofer · 2026-09-19
- [source] Jev, the decision-only model 20-200x faster than LLMs, spawns 359 projects — Loran750 · 2026-09-19
- Why Jev might finally kill the text prompt: millisecond decisions for generative GUIs — dpopa · 2026-09-20
- [source] Inside JEV: the $0.042/M-token fast-judgment model, black-box tested and dissected — AGI Hunt · 2026-09-20
- Jev hype gets pushback: embeddings plus logistic regression did this years ago — Loran750 · 2026-09-21
- JEV, a no-text decision model claims 200x speed and 400x cost savings, with an early port of Claude Code compaction hitting 4,575 stars in five days — gaganghotra_ · 2026-09-21
- ChatGPT co-creator's Jev: batch decisions without long LLM calls — DrDatta_AIIMS · 2026-09-21
3 near-duplicate retellings: multiply_matrix · multiply_matrix · hardimanjames