Tier 0 Proprietary service
What the publisher claims: System One model, type-safe, frontier
The one-liner: The open interface around a closed intelligence.
The six questions
No downloadable artifacts for the model itself; access is through the hosted API in early access.
No license grant; terms of use govern a service.
Runs only on TypeSafe's hosted service (70-500ms end-to-end, their infrastructure).
Workflow evals are published with full queries and methodology (an unusually candid disclosure of evaluation), but architecture, training data, and method (RLCD) are unpublished.
No path to reproduce: no weights, no paper, no architecture details.
Public SDK repos accept issues and the evals are published for scrutiny, but the model's development is entirely corporate; no community channel.
License
None published for the model — hosted service (early access) — license source ↗
- No weights, no architecture publication, no technical paper, no license
- Hosted only via console/API (also listed on OpenRouter as typesafe/jev-1.13 and jev-1.14)
- Python and JavaScript SDKs and the workflow-eval harness ARE open source — the interface is open, the model is not
What is actually released
- hosted API (early access)
- open-source Python/JS SDKs (github.com/typesafe-ai/system-one-adapter-python)
- workflow evals site (evals.typesafe.ai) with full queries and workflows
- no weights, no training data, no paper
Why it matters
The sharpest new form of the book's question: a decision model whose SDKs, eval harness, and type guarantees are all inspectable while the model itself is a black-box service. 'Open source' is doing zero work here — the openness lives entirely at the boundary.
Sources
- https://typesafe.ai/blog/introducing-system-one-models-and-jev
- https://en.wikipedia.org/wiki/Jev_(AI_model)
- https://github.com/typesafe-ai/system-one-adapter-python
- https://openrouter.ai/typesafe/jev-1.13
Derived from the openness dataset behind But Is It Open? — snapshot 2026-11-07 ↗. Scores are qualitative readings of the cited evidence, not verdicts; tier measures openness, not intelligence.