TypeSafe AI is the San Francisco company that makes Jev, the first model it calls a “System One” model. It was founded in 2024, spent about two years in stealth, launched Jev on 15 September 2026 with a $40 million seed round, and announced an $870 million Series A at a $7.5 billion valuation on 9 October 2026. This page covers who runs it, what it has shipped, which of its claims are independently checked, and how to tell its official products from the third-party ones that use the Jev name.
TypeSafe AI at a glance
| What it makes | “System One” models, which TypeSafe defines as models “built to make fast, structured decisions that software can use directly”. Jev is the first. |
|---|---|
| Founded | 2024 (Forbes, Bloomberg). The launch post mentions “two years in stealth”. |
| Headquarters | San Francisco, near the Embarcadero |
| Founders | Diogo Almeida (CEO), Sasha Sheng (COO), Erik Gafni (CTO) |
| Funding | $40M seed (September 2026); $870M Series A at $7.5B (October 2026) |
| Investors named | DCVC (seed lead), Andreessen Horowitz (Series A lead), Sequoia Capital, unnamed angels |
| Board | Martin Casado (a16z) joined with the Series A |
| Main product | Jev 1.13 (jev-1.13.0), $0.042 per million input tokens, output free |
Founders and team
TypeSafe’s team page lists three founders:
- Diogo Almeida, CEO. Signs the launch post as “founder” and posts on X as
@CompleteSkeptic. At OpenAI he worked on the instruction-following research that became InstructGPT and, later, ChatGPT; before that he was at Google Brain. Forbes reports he spent four years at OpenAI and left in 2024 to start TypeSafe. The company and its investors describe him as a co-inventor of RLHF and ChatGPT; that is their framing of a large team effort. - Sasha Sheng, COO. Former research engineer at Meta FAIR, with papers at NeurIPS and ECCV.
- Erik Gafni, CTO. A repeat founder, and an early employee at Invitae and Freenome.
The team page describes a small, flat team from OpenAI, Google Brain, Meta FAIR, Stripe, Airbnb, Plaid and Docker, working in person in San Francisco. TypeSafe has not published a headcount. On 11 October 2026 its jobs board listed ten open roles, from model research to a community manager.
Timeline
| Date | Event |
|---|---|
| 2024 | TypeSafe founded; its GitHub organisation was created in May 2024. The company stays in stealth for about two years. |
| 15 September 2026 | Launch. TypeSafe comes out of stealth with Jev “in early access”, a waitlist, the /v1/systemone API, Python and JavaScript SDKs, and a $40M seed round led by DCVC. Vercel AI Gateway lists Jev the same day. |
| 16 September | Vercel publishes its own AI SDK provider, @ai-sdk/typesafe-ai. |
| 21 September | Almeida’s Latent Space interview. He says the company had “almost no revenue before launch”. |
| Late September | The waitlist is reported to have been removed; TypeSafe’s docs now describe signing up and creating keys yourself. TypeSafe publishes an official n8n node (npm package created 29 September). |
| 30 September – 10 October | Competing decision models arrive from OpenAI, Cloudflare, Microsoft, Perplexity and others. See Jev alternatives. |
| 9 October | Series A: $870M at a $7.5B valuation, led by a16z, announced in a post titled “TypeSafe A raises Series AI”. Python SDK 0.7.4 released the same day. |
Funding
Seed: $40 million, September 2026
DCVC says it led the $40 million seed round announced at launch. Forbes reported a $200 million valuation, citing a person familiar with the deal; neither TypeSafe nor DCVC has confirmed that figure.
Series A: $870 million at $7.5 billion, October 2026
TypeSafe’s Series A post puts the round at “$870 million at a $7.5B valuation, with Martin Casado joining the board”. It was led by Andreessen Horowitz, with Sequoia Capital, existing investor DCVC and a group of angel investors. The post does not say whether the valuation is pre- or post-money. Bloomberg reported that a16z wrote the largest cheque. Plans for the money are stated in general terms: more “machine-native” models, infrastructure, and the enterprise features customers have asked for.
Before the round closed, The Information reported talks at a valuation above $10 billion. The round TypeSafe announced was smaller.
Company claims, and what backs them
Jev’s first four weeks produced a lot of large numbers. Most come from TypeSafe or its investors and have not been checked independently.
| Claim | Who says it | Status |
|---|---|---|
| Fortune 500 adoption | TypeSafe blog: “a third”. Almeida on X: 29.4%. a16z: 25%. | Company and investor claims. The three figures differ, and no customers are named. Bloomberg reports TypeSafe declined to name any. |
| Over $100M ARR | Latent Space AINews (10 October), citing a remark attributed to Sequoia’s Pat Grady | Second-hand. TypeSafe has published no revenue figure. |
| One trillion tokens in three days after launch | a16z | Investor claim |
| Over one million users within days | Bloomberg, reporting what the company said | Company claim |
| Up to 193.6× faster and 444.6× cheaper than LLMs | TypeSafe launch post | Vendor benchmark; TypeSafe itself calls these “on the higher end”. See our benchmark for independent numbers. |
| Jev “can’t hallucinate” | TypeSafe launch post | True of the output format (answers are always one of your options), not of correctness. See what Jev does. |
What is independently visible: Vercel told Fortune that nearly 13% of its paid teams tried Jev within a day of launch, and Jev ranks near the top of both third-party leaderboards, which we cover in Jev alternatives.
Official products
Everything TypeSafe publishes lives on the typesafe.ai domain, in the typesafe-ai GitHub organisation, or under the @typesafe-ai npm scope.
| Product | What it is | Where |
|---|---|---|
| Jev 1.13 | The model. Model ID jev-1.13.0; jev-latest and jev-preview both point to it. Text only, 32,000 tokens of state, $0.042 per million input tokens. | POST https://api.typesafe.ai/v1/systemone |
| Console | Playground and API keys | console.typesafe.ai |
| Docs | API reference, concepts, limits and pricing | docs.typesafe.ai |
| Python SDK | pip install typesafe-sdk, version 0.7.4 (9 October) | typesafe-sdk-python |
| JavaScript SDK | npm install @typesafe-ai/sdk, version 0.6.0 | typesafe-sdk-js |
| n8n node | Community node for n8n workflows. The official package is scoped: @typesafe-ai/n8n-nodes-typesafe-ai. An unscoped package with a similar name is from a different publisher. | n8n-nodes-typesafe-ai |
| Agent skill | Teaches coding agents to write Jev code | skills |
| System One adapter | The same client interface answered by an ordinary LLM, for side-by-side comparisons | system-one-adapter-python |
| WorkflowEvals | TypeSafe’s four-workflow benchmark (security incidents, agent traces, invoices, customer service), with code under Apache-2.0 | evals.typesafe.ai |
TypeSafe does not publish model weights, has no official MCP server, and publishes no models on Hugging Face. For what is and isn’t open, see Is Jev open source?
Official, partner or third-party?
Jev’s popularity brought a wave of services with “Jev” in the name. They fall into three groups.
- Official: TypeSafe’s own API, console, SDKs, n8n node and skill, listed above.
- Gateways that resell Jev at list price: TypeSafe’s Python SDK docs show how to reach Jev through OpenRouter (
typesafe/jev-1.13), Vercel AI Gateway (typesafe-ai/jev) and Pydantic AI Gateway. Cloudflare also liststypesafe/jevin its model catalog, marked “third-party”, though TypeSafe’s docs don’t mention it. All of them charge TypeSafe’s $0.042 per million input tokens. Vercel also maintains the@ai-sdk/typesafe-aiprovider; it is Vercel’s package, not TypeSafe’s. - Unaffiliated: independent APIs, MCP servers and model uploads that use the Jev name. Some are useful community projects; others resell access at a markup or are tied to a crypto token. Our guide to unofficial Jev APIs lists the ones we found and what to check.
For how Jev works under the hood, and what TypeSafe has and hasn’t said about its training method, read Jev architecture and RLCD. For what the model does in practice, start with what Jev does and when to use it.
Frequently asked questions
What is TypeSafe AI?
TypeSafe AI is a San Francisco startup, founded in 2024, that builds what it calls System One models: models that return typed decisions (yes/no probabilities, choices and scores) instead of generated text. Its first public model is Jev, launched on 15 September 2026.
Who founded TypeSafe AI?
TypeSafe's team page lists three founders: Diogo Almeida (CEO), Sasha Sheng (COO) and Erik Gafni (CTO). Almeida previously worked at OpenAI on the instruction-following research behind ChatGPT and at Google Brain; Sheng was a research engineer at Meta FAIR; Gafni is a repeat founder and was an early employee at Invitae and Freenome.
How much has TypeSafe AI raised?
A $40 million seed round led by DCVC, announced at launch on 15 September 2026, and an $870 million Series A at a $7.5 billion valuation, announced on 9 October 2026. Andreessen Horowitz led the Series A, with Sequoia Capital, DCVC and angel investors participating, and Martin Casado of a16z joined the board.
Does a third of the Fortune 500 use Jev?
That is TypeSafe's own claim, and it has not been independently checked. The company blog says a third, CEO Diogo Almeida posted 29.4%, and a16z's investment post says 25%. TypeSafe has not named the companies or said what counts as using Jev.
What is TypeSafe AI's revenue?
TypeSafe has not published revenue figures. Latent Space reported that the company passed $100 million in annual recurring revenue, citing a remark attributed to Sequoia's Pat Grady. Treat that as unconfirmed.
Is jevai.tools run by TypeSafe AI?
No. jevai.tools is an independent site with no affiliation with TypeSafe AI. Official TypeSafe sites are on the typesafe.ai domain: typesafe.ai, docs.typesafe.ai, api.typesafe.ai, console.typesafe.ai and evals.typesafe.ai.
Sources
- TypeSafe AI Introducing System One Models & Jev (launch post) ↗
- TypeSafe AI TypeSafe A raises Series AI ↗
- TypeSafe AI Team ↗
- TypeSafe docs Models: Jev 1.13, limits and pricing ↗
- TypeSafe docs SDKs ↗
- GitHub typesafe-ai organisation ↗
- TypeSafe AI WorkflowEvals ↗
- DCVC TypeSafe emerges from stealth with a new way of doing AI (15 September 2026) ↗
- a16z Investing in TypeSafe AI (9 October 2026) ↗
- X Diogo Almeida on the Series A (9 October 2026) ↗
- Forbes This $200 million startup wants to fix AI’s overconfidence problem (15 September 2026) ↗
- Fortune Jev has been a viral hit in Silicon Valley, but OpenAI is hot on its heels (8 October 2026) ↗
- Investing.com TypeSafe AI raises $870 million led by Andreessen Horowitz (Bloomberg report) ↗
- Latent Space Podcast interview with Diogo Almeida (21 September 2026) ↗
- Latent Space AINews: TypeSafe/Jev at >$100M ARR, $7.5B valuation (10 October 2026) ↗
- Vercel AI Gateway models API (typesafe-ai/jev) ↗
- TypeSafe docs Python SDK usage: connecting through AI gateways ↗
- OpenRouter TypeSafe: Jev model page ↗
- Cloudflare Model catalog: typesafe/jev ↗
- npm @typesafe-ai/n8n-nodes-typesafe-ai ↗
- npm @ai-sdk/typesafe-ai (published by Vercel) ↗
Checked on 11 October 2026. TypeSafe’s blog posts carry no publication dates, so launch and funding dates come from the investors’ and reporters’ dated posts. Figures that only TypeSafe or its investors have stated are labelled as their claims. This site is not affiliated with TypeSafe AI.
