Independent project. Not a U.S. government website.

USASI

Organization

TypeSafe AI

Model developer · Developer platform · San Francisco, California

TypeSafe AI is a San Francisco company that develops what it calls System One models, AI models that answer typed questions about supplied input with structured values and probabilities rather than generated text. Its first such model, Jev, launched in early access in September 2026 through a hosted API.453Fact reviewed Oct 1, 2026

Last reviewedEntry updated

Using this organization’s offerings

Ownership, delivery, and openness answer different questions. Publicly traded does not mean open source; privately held does not mean closed.

Intended users

Developers building software that needs a model's judgment as structured input to code, such as choosing an option, scoring against a rubric, or answering yes/no questions inside automated workflows.5Fact reviewed Oct 1, 2026

Hosted and self-hosted access

Jev is available through the hosted System One API, a console playground, and the Python and JavaScript SDKs. No model weights are published. The WorkflowEvals code and its Hugging Face datasets are public.6971314Fact reviewed Oct 1, 2026

Restrictions and review limits

Jev returns typed answers and probabilities, not generated text. The documentation lists known weaknesses with arithmetic, counting, dates, and multi-step indirection, says English is the primary training language, and warns that rate limits may change without notice.87Fact reviewed Oct 1, 2026

Key facts

Legal name
TypeSafe AI, Inc.1
Ownership (separate from openness)
Privately held

TypeSafe's team page says the company is backed by investors and offers employees equity. The pages read do not describe any public listing.3Fact reviewed Oct 1, 2026

Legal form
Unknown
Headquarters
San Francisco, California13
Founded
Unknown
Sectors
Research, Enterprise

What is and isn’t public

An evidence-based summary of what this catalog has documented. It is not a verdict on the organization as a whole.

TypeSafe offers Jev only through its hosted API. Its documentation describes one set of weights serving all accounts, and its Hugging Face organization lists datasets but no models. TypeSafe publishes its Python and JavaScript client SDKs on GitHub under the MIT License, and its WorkflowEvals harness code under Apache 2.0 (see the WorkflowEvals record).71312101114Fact reviewed Oct 1, 2026

Products

Documented products and how they are delivered. Each row links to the official page.

Hosted model products

Labeled Hosted product/API — this describes how a documented product is delivered (hosted access). It does not determine whether underlying weights or code are open; assess the specific release and its licenses.

TypeSafe AI: hosted model products
ProductKindDocumented delivery and accessReviewed
System One API (Jev) (external site: docs.typesafe.ai)Hosted model APIHosted API (POST /v1/systemone) authenticated with API keys from the TypeSafe console, which also offers a browser playground. The current model, Jev 1.13, accepts text input only and is priced per input token. TypeSafe opened early access with a developer waitlist at launch.764Oct 1, 2026

Other products

TypeSafe AI: other products
ProductKindDocumented delivery and accessReviewed
TypeSafe client SDKs (external site: docs.typesafe.ai)Developer toolPython and JavaScript/TypeScript client libraries for the System One API, published on GitHub under the MIT License; the Python SDK installs as the typesafe-sdk package.961011Oct 1, 2026

Cataloged artifacts

Records in the Open Models & Tools directory that list this organization as a maintainer or publisher.

Evaluation tools (1)

Notable documented facts

  • TypeSafe's documentation says Jev is trained with a method the company calls Reinforcement Learning for Calibrated Decisions (RLCD), is not fine-tuned on customer data, and serves every account from the same weights.74
  • TypeSafe publishes a list of known failure modes for jev-1.13, including literal reading of instructions, arithmetic and counting, date comparison, and text generation. For those tasks it advises doing the work in code or using a generative model instead.8
  • TypeSafe publishes its WorkflowEvals harness and datasets, which compare models on four automation workflows. Reference labels in these datasets come from large models of other providers, not from human verification.141513

Open positions

Automated job coverage is not configured for this organization.

What this catalog does not know

Unknown means the sources reviewed for this record do not document it. It is not evidence that something does not exist.
  • Founding date: not documented in the sources reviewed.

Have a primary source? How to report a correction.

U.S. eligibility

How this record meets the catalog’s published eligibility policy.

Eligible · basis: U.S. headquarters

TypeSafe AI's terms of use (last updated September 19, 2026) state that the site is offered by TypeSafe AI, Inc., located at 255 California St, Suite 1300, San Francisco, California, and give the same address for legal notices. The team page says staff work in person at the company's San Francisco office, and the privacy policy says the services are hosted in the United States. The terms choose Delaware law but do not state where the company is incorporated; this assessment does not rely on an SEC filing or a state registry record.132

Assessed Oct 1, 2026

Eligibility policy

Sources

  1. 1.
    Terms of use - TypeSafe AI (external site: typesafe.ai)

    TypeSafe AI, Inc. · Official page · published Sep 19, 2026 · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  2. 2.
    Privacy policy - TypeSafe AI (external site: typesafe.ai)

    TypeSafe AI, Inc. · Official page · published Nov 19, 2025 · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  3. 3.
    Team - TypeSafe AI (external site: typesafe.ai)

    TypeSafe AI, Inc. · Official page · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  4. 4.
    Introducing System One Models & Jev - TypeSafe AI Blog (external site: typesafe.ai)

    TypeSafe AI, Inc. · Announcement · published Sep 15, 2026 · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  5. 5.
    Introduction - TypeSafe AI Documentation (external site: docs.typesafe.ai)

    TypeSafe AI, Inc. · Documentation · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  6. 6.
    Quick start - TypeSafe AI Documentation (external site: docs.typesafe.ai)

    TypeSafe AI, Inc. · Documentation · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  7. 7.
    Models - TypeSafe AI Documentation (external site: docs.typesafe.ai)

    TypeSafe AI, Inc. · Documentation · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  8. 8.
    Jev 1.13 jaggedness - TypeSafe AI Documentation (external site: docs.typesafe.ai)

    TypeSafe AI, Inc. · Documentation · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  9. 9.
    Client SDKs - TypeSafe AI Documentation (external site: docs.typesafe.ai)

    TypeSafe AI, Inc. · Documentation · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  10. 10.
    typesafe-sdk-python LICENSE (MIT) (external site: raw.githubusercontent.com)

    TypeSafe AI, Inc. (GitHub) · License · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  11. 11.
    typesafe-sdk-js LICENSE (MIT) (external site: raw.githubusercontent.com)

    TypeSafe AI, Inc. (GitHub) · License · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  12. 12.
    TypeSafe (typesafe-ai) on GitHub (external site: github.com)

    TypeSafe AI, Inc. (GitHub) · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  13. 13.
    TypeSafe AI on Hugging Face (external site: huggingface.co)

    TypeSafe AI, Inc. (Hugging Face) · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  14. 14.
    typesafe-ai/WorkflowEvals (GitHub repository) (external site: github.com)

    TypeSafe AI, Inc. (GitHub) · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

  15. 15.
    Workflow evals - TypeSafe AI (external site: evals.typesafe.ai)

    TypeSafe AI, Inc. · Official page · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

Support Us

Help keep USASI useful.

Optional. No USASI account required. Payment takes place on the linked provider’s website (Buy Me a Coffee).

About supporting this project