TypeSafe AI Raises $870 Million Series A at $7.5 Billion Valuation

TypeSafe AI, the San Francisco startup founded in 2024, closed an $870 million Series A round on October 9, 2026, at a $7.5 billion post-money valuation. Andreessen Horowitz led the round with participation from Sequoia Capital, DCVC, and angel investors. The announcement followed the September 15 launch of the company’s Jev decision model by 24 days.

TypeSafe AI funding round details and Jev model performance

The round values TypeSafe AI roughly 37 times higher than its $40 million seed round closed on September 15 at a $200 million post-money valuation. Andreessen Horowitz general partner Martin Casado joined the board. Company data shows the Jev model processes unstructured input and developer-defined typed questions to return one of three structured outputs: Choice from up to 255 options, Score on a numeric scale, or Noul binary probability, each accompanied by calibrated confidence scores.

Jev operates as a transformer-based System One architecture that avoids text generation. Reported latency ranges from 70 to 500 milliseconds end-to-end. Pricing stands at $0.042 per million input tokens with free output tokens. Internal benchmarks cited by TypeSafe claim 100- to 200-fold improvements in speed and cost versus frontier large language models on classification and routing tasks. The model supports parallel question processing and targets high-volume, low-latency software decisions such as triage, classification, and routing.

Adoption metrics reported by the company include more than one million users shortly after launch and one trillion tokens processed in the first three days. Roughly one-third of Fortune 500 companies have used the service according to TypeSafe statements. A recruiting platform case study reported an 88 percent cost reduction compared with Gemini 3.1 Flash Lite.

Background on founders and competition in decision models

CEO Diogo Almeida previously worked at OpenAI as fourth author on the InstructGPT paper and contributed to RLHF methods that supported ChatGPT. CTO Erik Gafni and COO Sasha Sheng, formerly of Meta FAIR, complete the founding team. The company spent approximately two years in stealth developing the parallel sampler and RLCD training method before releasing Jev.

OpenAI introduced a competing Decisions API built on GPT-6 Luna on October 6, 2026, at $0.10 per million input tokens. Microsoft released a matching offering at the same price point. The launches established a distinct decision-model category separate from general-purpose language models. TypeSafe positions Jev for enterprise workloads that require deterministic, auditable outputs rather than open-ended text.

Funds will support expansion of the System One model family, addition of enterprise security and governance features, and scaling of infrastructure. No revenue or forward guidance figures have been disclosed beyond customer-reported savings.

Enterprise implications and next steps for decision model adoption

Software teams evaluating decision models can now compare latency, pricing, and output calibration across TypeSafe, OpenAI, and Microsoft offerings. Early adopters report replacing multiple LLM calls with single Jev queries for routing and classification logic, which reduces both cost and latency variance. The rapid valuation increase from seed to Series A reflects investor interest in specialized architectures that deliver measurable efficiency gains on high-volume inference tasks.

Enterprises integrating these models should test calibration accuracy on internal data sets and review governance controls before production deployment. TypeSafe plans to release additional model variants and expanded enterprise tooling in the coming months. Observers will watch whether independent benchmarks confirm the claimed performance advantages and whether the new category sustains separate pricing from general language models.