TypeSafe must meter 23.8 trillion input tokens to bill a million dollars for Jev
Jev charges $0.042 per million input tokens and nothing for output, so TypeSafe AI's revenue moves only with the state that agents pass in. Vercel, Cloudflare, LangChain and Langfuse listed it within a week.
The Investor · Invest desk
What happened
- TypeSafe AI launched Jev on September 15 as a transformer model that returns probabilities instead of text, answering questions such as which tool to call next in under half a second for $0.042 per million input tokens, with output free.
- Within days of launch, Vercel, Cloudflare, LangChain and Langfuse had added Jev to their model gateway stacks.
- TypeSafe dropped the waitlist on September 21 and opened Jev to everyone, with access starting at $5 in credits, which the company says is equivalent to roughly 120 million tokens.
- Vercel's listing describes Jev as a System One evaluation model that scores typed questions against shared state and returns choices, scores and boolean probabilities, with many questions evaluated in one request.
- One engineer swapped OpenAI's GPT-5.6 Luna for Jev in a command-safety classifier and got results five to 18 times quicker, while another found Gemini slightly more accurate on email classification and Jev about 10 to 20 times cheaper.
Why it matters
- What it limits constraint Free output leaves TypeSafe one billable variable, the size of the state a developer passes in, so revenue per decision falls every time a customer trims context to save money.
- What doesn't add up contradiction Vercel prices the model at Free on its own gateway page while TypeSafe quotes $0.042 per million input tokens, so the four-platform pickup measures shelf space.
- What it costs cost The saving from replacing a metered language-model call with free output lands in the agent vendor's margin, and TypeSafe collects only on the context handed to the model.
- Who has to decide decision Every team running classification inside an agent now has a priced choice, accepting slightly lower accuracy per call in exchange for an order-of-magnitude cost gap on all of them.
Next story · Build · Oct 9
Andreessen Horowitz leads a reported $870M round for Jev maker TypeSafeConfirmed7 publishers
At $0.042 per million input tokens, a million dollars of billings means 23.8 trillion tokens through the meter [2][21]. A million routing calls carrying 2,000 tokens of state each bills $84 [19]. Raise the state to 20,000 tokens for a rubric evaluation and the same million calls bills $840 [20]. The $5 entry price is consistent with input-only metering: $5 divided by $0.042 per million is 119 million tokens, near the roughly 120 million TypeSafe quotes [22].
The record on demand is thinner than the record on distribution. Surging demand last week briefly cost TypeSafe the ability to serve users from its API, according to the Indian Express [17]. Neither that account nor Vercel's model page puts a number on revenue, paying customers or funding [12]. TypeSafe has kept Jev's architecture under wraps, and the Indian Express reports that some observers speculate it is built on top of an open-weight LLM [16].
Diogo Almeida invented reinforcement learning from human feedback and was behind the instruction-following that underpins ChatGPT [24]. He left OpenAI nearly two years ago and founded TypeSafe AI with Erik Gafni and Sashsa Sheng [8]. "The problem is we are optimising for human language. We have been super good at human language for four years, but it's not useful for automation because computers speak a different language," he said [11]. Jev was trained only on synthetically generated data, using a method the company calls Reinforcement Learning for Calibrated Decisions [15], and TypeSafe says the model cannot hallucinate because the user supplies the context and defines the outputs in advance [23].
The company named the model after William Stanley Jevons, whose paradox holds that cheaper coal leads to more coal being consumed [14]. "We think that there's just going to be smart software all over the place in a way that's emergent and distributed ... much more like the early internet than you know like the mega apps that people are trying to build right now," Almeida was quoted as saying by TechCrunch [13].
In my view the pricing is the main risk. The buyers the Indian Express identifies are agent vendors switching in order to spend less [25], output is free [2], and the volume that makes input-only metering work has not been shown. The case on the other side is the company's own name: if each agent task fires dozens of cheap decisions, 23.8 trillion tokens is a year of traffic, not a decade [21]. And if the accuracy gap one developer found against Gemini matters more on a safety or routing call than the price gap does [10], the frontier model stays in the loop and Jev stays a small line item. TypeSafe says it will build more versions of the model in new modalities [18].
What to watch
- A gateway moving Jev from a Free listing to a metered rate would be the first public sign of paid volume.
- Any disclosure from TypeSafe of billed token volume, customer numbers, or a funding round.
- Published head-to-head accuracy tests against Gemini on routing and classification tasks.
Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence50
- Adoption35
- Hype gap+35
- Incentives50
- Confidence45
Perspective Coverage
3 publishers- Builder
- Builder 53%
- Operator
- Operator 27%
- Investor
- Investor 20%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Jev is not a chatbot or text-output large language model; the transformer-based model produces probabilities or 'calibrated decisions', according to the startup.
- [2]
Jev can help AI agents decide which tool to call next and whether to retry a task, answering those questions in under half a second for $0.042 per million input tokens, with output free.
- [3]
A few days after launch, Jev had been added by model gateway platforms including Vercel, Cloudflare, LangChain and Langfuse to their stacks.
- [4]
Vercel describes Jev as TypeSafe AI's System One evaluation model for fast, structured decisions in software, which evaluates shared state against typed questions and returns choices, scores and boolean probabilities, supporting classification, routing, rubric-based assessment and automated verification, with multiple questions evaluated in parallel within a single request.
- [5]
TypeSafe AI's model Jev was launched on September 15.
- [6]
TypeSafe AI announced on Monday, September 21 that Jev is available to everyone without a waitlist, with access starting at $5 in credits, equivalent to roughly 120 million tokens.
- [7]
Vercel's AI Gateway page for Jev lists the price as Free.
- [8]
Almeida exited OpenAI nearly two years ago and launched TypeSafe AI with co-founders Erik Gafni and Sashsa Sheng.
- [9]
One software engineer replaced OpenAI's GPT-5.6 Luna with Jev to run a classifier reviewing commands for safety, and the results were five to 18 times quicker with greater accuracy.
- [10]
A developer comparing Jev with Google's Gemini models on tasks such as classifying business emails found Gemini slightly more accurate while Jev was about 10 to 20 times cheaper.
- [11]
"The problem is we are optimising for human language. We have been super good at human language for four years, but it's not useful for automation because computers speak a different language," Almeida said.
- [12]
Neither the Indian Express account nor Vercel's model page states a revenue figure, a paid customer count or a funding amount for TypeSafe AI.
- [13]
"We think that there's just going to be smart software all over the place in a way that's emergent and distributed ... much more like the early internet than you know like the mega apps that people are trying to build right now," Almeida was quoted as saying by TechCrunch.
- [14]
Jev is named after William Stanley Jevons, the economist behind the paradox that cheaper coal will lead to more coal consumption; Jevons's paradox has often been cited by tech industry leaders such as Microsoft CEO Satya Nadella, who expect intelligence to follow the same curve.
- [15]
Jev was trained exclusively on synthetically generated data using a method TypeSafe AI calls Reinforcement Learning for Calibrated Decisions.
- [16]
TypeSafe AI has kept the finer details of Jev's model architecture under wraps, though some observers speculate it has been built on top of an open-weight LLM.
- [17]
Surging demand last week caused TypeSafe to lose the ability to serve users from its API for a short period.
- [18]
TypeSafe has said it will be building more versions of the model in new modalities.
- [19]
A million Jev calls each passing 2,000 tokens of state bills $84 at the launch input rate.
- [20]
A million Jev calls each passing 20,000 tokens of state bills $840 at the launch input rate.
- [21]
Billing $1 million at $0.042 per million input tokens requires about 23.8 trillion input tokens.
- [22]
$5 of credits buys about 119 million input tokens at $0.042 per million, close to the roughly 120 million tokens TypeSafe quotes, which is consistent with input-only metering.
- [23]
TypeSafe claims Jev is a first-of-its-kind 'System One' model focused on intuition rather than reasoning, and says it is incapable of hallucinating because users give it context and define the outputs in advance.
- [24]
Diogo Almeida is a researcher whose work enabled the instruction-following feature underpinning OpenAI's ChatGPT, and he invented reinforcement learning from human feedback (RLHF), a post-training technique that drove the scaling of LLMs.
- [25]
Agent vendors are likely to be the target customers, as companies running multiple agents relying on costly language-model calls could replace them with a model that offers free output, potentially lowering the cost of running each agent and improving their margins.
Sources
3 independent publishers whose own reporting we read for this story.
- cryptobriefing.comTypeSafe opens Jev AI to public after rapid adoption forces waitlist removal
1 article · September 21, 2026
- indianexpress.comMeet Jev, the new AI model built to make decisions instead of generating text | Technology News - The Indian Express
1 article · September 21, 2026
- Jev API, Pricing & Playground
vercel.com
1 article · September 21, 2026