Pomegra
TypeSafe AI raised an $870M Series A led by a16z at a $7.5B valuation, 24 days after leaving stealth with a $40M seed to build decision models for software.
Key Takeaways
- TypeSafe AI raised $870M at a $7.5B valuation, led by a16z, on October 9, 2026.
- The round came 24 days after a $40M seed at about $200M post-money.
- Its model, Jev, returns typed decisions instead of text; adoption claims are self-reported.
Lead
TypeSafe AI, a San Francisco AI lab, announced an $870 million Series A on October 9, 2026, at a $7.5 billion valuation. Andreessen Horowitz led the round, with Sequoia Capital and DCVC participating. Martin Casado of a16z joins the board.
The raise arrives 24 days after TypeSafe left stealth on September 15 with a $40 million seed led by DCVC at roughly $200 million post-money. If those two valuations are comparable, the step-up is about 37.5 times in under a month.
What Does TypeSafe AI Actually Build?
TypeSafe builds machine-native models meant to run as components inside software rather than as chat assistants. Its first release, Jev, is what the company calls a "System One" model. It is a transformer-based decision model that outputs a probability, a yes/no answer, a numerical score, or a choice from a list the developer defines. It produces no natural language.
The company says Jev responds in 70 to 500 milliseconds, against 3 to 329 seconds for frontier language models on comparable queries. Published pricing is $0.042 per million input tokens, and the context window is 32,000 tokens. Typical uses include checking agent tool calls, scoring security alerts, triaging support tickets, flagging fraud and judging other models' output.
Founder and CEO Diogo Almeida previously worked at OpenAI, where he contributed to InstructGPT, ChatGPT and GPT-4. Erik Gafni is CTO and Sasha Sheng is COO. The company spent about two years in stealth before launch.
How Much Adoption Backs a $7.5 Billion Price?
The adoption figures are large, but they come from the companies involved. TypeSafe says about one-third of the Fortune 500 uses Jev, while a16z's own figure is 25%. a16z also cites roughly 1 trillion tokens processed within days of launch, and reports put Jev on 13% of Vercel's paid teams within 24 hours. Jev is available through Vercel AI Gateway, OpenRouter, LangChain, Pydantic AI and DSPy.
No independent benchmarks accompanied the round. TypeSafe's speed and cost comparisons pit a specialized decision model against general-purpose language models, which are built for a broader job. An outside analysis cited in coverage found a median 7x speedup and 30x cost savings, but the methodology behind those numbers has not been widely reviewed.
The model also has limits. Version 1.13 shows unreliable counting, arithmetic and date comparison, and a typed output can be wrong while still being valid. The "hallucination-free" label applies to output format, not to factual accuracy.
Why Is the Valuation Jump So Steep?
The step-up reflects a bet on a new model category, not on revenue. TypeSafe has not disclosed revenue, contract values or customer counts beyond usage claims. A 37.5x increase in 24 days is hard to read as a repricing of the same business. It looks more like a second, much larger bet on the same thesis: that software will call fast, cheap decision models thousands of times per workflow.
Sequoia and DCVC held positions or joined in the round, so the new money did not displace the seed backers. The size of the cheque, $870 million for a company a few weeks out of stealth, also signals how much capital is chasing infrastructure for AI agents.
Who Else Is Competing in Decision Models?
Competition arrived within days of launch. OpenAI introduced a Decisions API on October 6 at $0.10 per million tokens. Microsoft released Microsoft-Decision-1 on October 9, priced at $0.042 per million tokens, the same as Jev. Perplexity, AWS and Cloudflare have also shipped decision models, according to coverage of the round.
Matching prices from a company with Microsoft's distribution narrows the cost argument. TypeSafe's defense will rest on accuracy, latency and how deeply Jev embeds in developer tooling.
What Will TypeSafe Do With the Money?
TypeSafe plans to extend the System One series beyond Jev and add enterprise features. It also plans to expand into image, audio, video and sensor data. Each of those moves requires training new models, which explains the size of the raise.
Outlook
TypeSafe now has $870 million, a $7.5 billion valuation and a product with fast reported adoption but no independent verification. The next tests are whether usage converts into disclosed revenue, whether Jev holds its accuracy claims under outside benchmarking, and whether it keeps a price or performance edge against OpenAI and Microsoft. A new model class would need to survive all three to justify a valuation set 24 days after the seed.
