TypeSafe's Jev becomes the fastest-adopted model in Vercel AI Gateway history

TypeSafe's Jev reached 35% of Vercel AI Gateway's paid teams within two weeks of its September 15 launch, nearly double the reach of Sonnet 5.5, the second-most-used model on the gateway, according to Vercel's usage data. Jev, designed by TypeSafe AI co-founder and former OpenAI researcher Diogo Almeida to return only typed Booleans, Choices, or Scores with attached probabilities rather than prose, hit roughly 13% of paid teams within 24 hours; TypeSafe benchmarks put it at up to 193.6 times faster and 444.6 times cheaper than frontier LLMs on equivalent workflow tasks at $0.042 per million input tokens with free output. Vercel, Cloudflare, LangChain and Langfuse integrated Jev within three days of launch, and TypeSafe has raised a $40 million seed led by DCVC plus $870 million more at a $7.5 billion valuation.

TypeSafe's Jev reached 35% of Vercel AI Gateway's paid teams within two weeks of launch, nearly double the next closest model, by answering questions instead of writing essays. Diogo Almeida did not build Jev to chat. The TypeSafe AI co-founder and former OpenAI researcher, who co-invented the RLHF technique behind ChatGPT, designed it to return a yes, a no, a pick from a list, or a score, nothing else. No prose, no formatting, no hallucinated caveats. Just a typed answer in under half a second. That narrowness turned out to be exactly what a huge slice of Vercel's developer base wanted. According to Vercel's own usage data, Jev hit roughly 13% of the AI Gateway's paid teams within 24 hours of its September 15 launch. Two weeks in, that number had climbed to 35%: nearly double the reach of Sonnet 5.5, the second-most-used model on the gateway, and by Vercel's count the fastest adoption curve any model has posted in the product's history. TechBytes and Sacra both cited the same internal figures in their coverage of the rollout. The mechanics explain the speed. Jev takes in application state - whatever a trading bot, support router, or coding agent is currently juggling - and returns a typed Boolean, Choice, or Score, each with an attached probability. A developer building an agent that needs to decide "approve this refund or escalate it" no longer has to send that decision to a frontier model, wait a second or two, then parse free text back into a usable value. TypeSafe's own benchmarks put Jev at up to 193.6 times faster and 444.6 times cheaper than frontier LLMs on equivalent workflow tasks, and it prices at $0.042 per million input tokens with output entirely free. Nobody was hyping Jev the way they hyped GPT-5.6 or Fable 5.1. It doesn't write code, doesn't pass bar exams, doesn't show up on leaderboards built for general intelligence. And that's precisely the point. Agentic products - coding assistants deciding which file to touch next, trading systems deciding whether to execute, support pipelines deciding whether to escalate - don't need a model that can write a sonnet. They need one that decides fast, cheaply, and at volume, thousands of times a minute if the workload demands it. TypeSafe's Jev becomes Vercel AI Gateway's fastest-adopted model ever https://startupfortune.com/typesafes-jev-becomes-vercel-ai-gateways-fastest-adopted-model-ever/ Vercel, Cloudflare, LangChain and Langfuse integrated Jev within three days of its September 15 launch, and TypeSafe's seed round came in at $40 million, led by DCVC, backing ex-OpenAI researcher Diogo Almeida. - typescript model fastest adoption vercel ai gateway https://startupfortune.com/typesafes-jev-becomes-vercel-ai-gateways-fastest-adopted-model-ever/ - jev production verdict api model adoption rate https://startupfortune.com/typesafes-jev-becomes-vercel-ai-gateways-fastest-adopted-model-ever/ That's a different purchasing decision than the one frontier labs have spent two years optimising for. TypeSafe raised a $40 million seed led by DCVC to build exactly this niche, then raised $870 million more at a $7.5 billion valuation: a staggering number for a company whose flagship product deliberately refuses to generate text. Investors aren't betting on Jev replacing ChatGPT. They're betting that every agentic pipeline running today has dozens of small, repeated decisions buried inside it, and that routing those decisions to a $0.042-per-million-token specialist beats paying frontier-model rates for the same yes-or-no answer. The gateway is the one making the call None of this reaches a developer's code without Vercel's AI Gateway sitting in between, and that's the second story here. Gateways like Vercel's, and OpenRouter, which Stripe agreed to acquire for more than $7.5 billion according to reporting from Axios and the New York Times, have quietly become the layer that decides which models actually get used in production. A model can top every public benchmark and still go nowhere if it never gets slotted into the handful of routing products developers already have wired into their stacks. Stripe's OpenRouter deal makes the same point from the money side. OpenRouter already routes more than 400 models from over 60 providers through a single API serving some 8 million users, and pairing it with Stripe's earlier purchase of the usage-metering firm Metronome gives Stripe control over both which model answers a request and how that usage gets billed. Vercel doesn't have Stripe's billing muscle, but the Jev numbers show the same dynamic at work: a gateway's own adoption dashboard is becoming a more useful signal of where developer demand is actually headed than any benchmark chart a lab publishes. Frontier labs are still going to keep chasing bigger context windows and higher eval scores, and that race isn't going away. But Jev's climb suggests a second market has opened up underneath it, one where the product being sold isn't intelligence, it's latency. For a founder building an agent that fires off a few thousand small decisions an hour, that tradeoff is not a close call. Also read: Disney hires the Character.AI boss it once threatened to sue https://startupfortune.com/disney-hires-the-characterai-boss-it-once-threatened-to-sue/ • India's cyber agency says AI agents read confidential files without permission https://startupfortune.com/indias-cyber-agency-says-ai-agents-read-confidential-files-without-permission/ • Anthropic warns AI models could soon copy themselves onto smart devices https://startupfortune.com/anthropic-warns-ai-models-could-soon-copy-themselves-onto-smart-devices/ This article is posted in AI News https://startupfortune.com/category/ai/ , check it out for more related stories. Join the discussion Open in the community → https://startupfortune.com/community/ Almost there. Sign in and your reply posts straight away.