TypeSafe’s Jev Now Serves 31% of OpenRouter Classification

TypeSafe’s Jev served 31% of the classification requests made on OpenRouter in the week of 9 November, according to an OpenRouter chart that Peter Walker, OpenRouter’s head of insights, posted on X on 17 November. The chart’s title reads “Jev now serves nearly 1 in 3 classification requests”. Jev was first listed on OpenRouter on 18 September.

Our number: in the week before the chart marks Jev’s launch, Google and OpenAI together held about 47.8% of those requests, by our reading of the chart. In the week of 9 November they held 26%, 11% for Google and 15% for OpenAI, a drop of about 22 percentage points, by our arithmetic.

Business Pill · STRUCTURED OUTPUT

A short Business Engineer explainer of structured output: getting a model to answer in a fixed, machine-readable form. It teaches the general idea only and does not discuss TypeSafe, Jev or OpenRouter’s chart.

The key insight: As we read it, OpenRouter’s chart shows a narrow job changing suppliers fast: a model from a company new to OpenRouter in September went from no classification share to 31% in its eighth week on the chart, while Google and OpenAI together fell from about 47.8% to 26%, by our reading and arithmetic.

What OpenRouter’s Chart Shows

Walker wrote with the chart: “It is possible, even today, to create a new AI company and with your first release take real, meaningful share away from the entire existing model ecosystem.” He added: “May we continue to see more Jevs!”

The chart is headed “Classification requests by lab, weekly | Jul 6 – Nov 15, 2026”. It shows 19 weekly bars, each split by lab, and its footer reads “Source: OpenRouter analytics”. A marker reading “Jev launches” sits over the week of 21 September; there is no Jev segment before it.

For the latest week the chart labels Jev at 31%, OpenAI at 15%, DeepSeek at 14%, Google at 11%, Qwen at 5% and “53 other labs” at 24%. By our reading of the bars, Jev held about 19.4% in its launch week, about 27.8% a week later and about 31.3% by the week of 5 October. It peaked at about 34% in the weeks of 26 October and 2 November. That makes eight weekly bars from launch to the latest week.

Share of weekly classification requests on OpenRouter for Jev, OpenAI and Google, from the OpenRouter analytic
Share of weekly classification requests on OpenRouter for Jev, OpenAI and Google, from the OpenRouter analytics chart Peter Walker posted on X on 17 November 2026. Weekly values are our readings of the chart (approximate); the end values are OpenRouter’s labels.

What Was Posted on X

Posted by Peter Walker, head of insights at OpenRouter per his X bio, at 00:51 UTC on 17 November, this post says a new AI company can take “real, meaningful share” with its first release. Its chart is titled “Jev now serves nearly 1 in 3 classification requests” and credits OpenRouter analytics. The piece takes the 31% from that chart’s latest-week label.

What Jev Is

OpenRouter’s TypeSafe page describes Jev 1.14 as “the latest version of Jev, TypeSafe’s structured decision model and the first of its System One models.” It says System One models “make fast, structured decisions for software, returning a typed answer rather than free-form text”, and that Jev is “suited for routing, classification, and other decision points inside an application”.

OpenRouter lists Jev 1.13 from 18 September and Jev 1.14 from 14 October. Its Jev Latest page lists “$0.042 per million input tokens, $0 per million output tokens.” A Jev Router, listed on 25 September, “picks the best model and reasoning effort for each request” and runs on Jev, according to OpenRouter’s model listing.

TypeSafe said in October that it had raised an $870 million Series A at a $7.5 billion valuation, led by Andreessen Horowitz (our funding piece). Its launch claims of 193.6x faster and 444.6x cheaper than LLMs on System One tasks are the vendor’s own (our claims check).

Who Lost Share

Between the week of 14 September and the week of 9 November, Google went from about 25.5% to 11%, OpenAI from about 22.3% to 15% and DeepSeek from about 22.0% to 14%, by our reading of the chart against its labels. Qwen went from about 9.9% to 5%. The 53 other labs went from about 20.4% to 24%.

Google’s share was already falling before Jev: the chart puts it at about 35% in the week of 6 July, by our reading. The chart does not say why any lab’s share moved.

Rivals have since launched their own decision models. Google launched Gemini Decision 1 on 11 November at $0.10 per million input tokens (our Gemini Decision 1 piece). Perplexity released the weights of its Decider models, which OpenRouter is serving free until 19 November (our Perplexity Decider piece).

Dumbbell chart of OpenRouter classification request shares, week of 14 Sep vs week of 9 Nov 2026: Jev 0% to 31%, Google about 25.5% to 11%, OpenAI about 22.3% to 15%, DeepSeek about 22% to 14%, Qwen about 9.9% to 5%, 53 other labs about 20.4% to 24%
Shares of weekly classification requests on OpenRouter, week of 14 September (our readings of OpenRouter’s chart, approximate) against the week of 9 November (OpenRouter’s labels). The 22-point combined Google and OpenAI drop is our arithmetic.

How It Fits OpenRouter’s Wider Data

Classification is one slice of OpenRouter’s traffic. Across all text requests in the week beginning 9 November, DeepSeek leads at 31.8%, according to OpenRouter’s rankings page, and Jev is not among the nine authors it lists by name. OpenRouter’s task table, ranked by share of spend, lists both Jev versions under classification and tagging.

On 16 November Walker posted an OpenRouter chart showing agent tokens up 106x this year (our agent-token piece). DeepSeek V4.1 Flash led OpenRouter by tokens at 61.2 trillion in the week through 16 November (our DeepSeek piece), and Hugging Face is moving its chat models onto OpenRouter routing (our Hugging Face piece).

Foundation Capital’s Jaya Gupta has argued that “Once the application holds the state, every model becomes interchangeable” (our piece on her argument). Gamma’s own spend data showed OpenAI passing Anthropic, 40% to 38% (our Gamma piece).

The Structural Read

As we read the chart, the move came quickly and then held: Jev was at about 27.8% in its second week and has stayed between about 30% and 34% since the week of 5 October, by our reading of the bars.

As we read OpenRouter’s pages, Jev is built for this job: OpenRouter describes it as returning “a typed answer rather than free-form text”, and the Jev Latest page lists $0 per million output tokens.

Peter Walker, head of insights at OpenRouter, on X, 17 November 2026

“It is possible, even today, to create a new AI company and with your first release take real, meaningful share away from the entire existing model ecosystem.”

Three Implications

FOR MODEL LABS Google’s classification share went from about 25.5% the week before Jev launched to 11% in the week of 9 November, by our reading of OpenRouter’s chart.

FOR BUYERS OpenRouter lists Jev Latest at $0.042 per million input tokens and $0 per million output tokens; Google priced Gemini Decision 1 at $0.10 per million input tokens.

RELATED Our TypeSafe funding piece covers the $870 million Series A at a $7.5 billion valuation.

The Business Engineer Lens

This story maps onto the Business Engineer framework The State of the Inference Economy.

The framework says: “Inference is a revenue engine: continuous, distributed across millions of applications and agents, incurring marginal cost with every query.”

As we read it, classification is that marginal cost at its most repetitive: a short decision inside an application, made again on every request, and OpenRouter’s chart shows that job moving to a model built and priced for it. The framework does not discuss TypeSafe, Jev or this chart.

What Is Not Established

The chart gives shares, not request counts, so the number of classification requests behind the 31% is not public in what we read. It does not say how many customers send Jev’s requests, or what share of classification spend Jev takes. The weekly values other than the latest week’s labels are our readings of the bars, approximate to about one percentage point.

The chart does not explain how OpenRouter assigns a request to classification. We did not contact TypeSafe or OpenRouter for this piece.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The State of the Inference Economy

The Business Engineer analysis of how inference, not training, became the revenue engine of AI.

Read the premium analysis →Get 50% off →

The Bottom Line

OpenRouter’s own chart puts TypeSafe’s Jev at 31% of weekly classification requests in its eighth week on the chart, ahead of OpenAI’s 15% and Google’s 11%. By our reading, Google and OpenAI together fell from about 47.8% to 26% over that stretch.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. We read Peter Walker’s post on X of 17 November 2026 and the OpenRouter chart attached to it, and OpenRouter’s TypeSafe, Jev Latest and rankings pages on 17 November 2026. The latest-week shares are OpenRouter’s labels; earlier weekly values are our readings of the chart’s bars, approximate to about one percentage point. Walker’s role is from his X bio. Funding, competitor and other platform figures are from our earlier pieces, linked above. Combined shares and differences are our arithmetic. TypeSafe and OpenRouter were not contacted for this piece.

Sources: Peter Walker (OpenRouter) on X, 17 Nov 2026 · OpenRouter: TypeSafe models page · OpenRouter: Jev Latest page · OpenRouter: rankings

Scroll to Top
FourWeekMBA