TypeSafe API
TypeSafe AI builds System One models: models that answer typed questions with structured results instead of writing prose. Its flagship model, Jev, is available through Felo, so a decision layer can share the same key as your chat, image and video models.
- Models
- 1
- decisions ×1
- Input per 1M
- $0.042
- Output per 1M
- $0
- Largest context
- 262.144K
What TypeSafe provides
TypeSafe AI introduced the System One category and Jev, which it describes as the first System One model. A request carries a state plus typed questions, and the reply contains values your code can branch on: a choice with probabilities and confidence, a score on a rubric you define, a noul yes/no probability between 0 and 1, or a bounding box for an object in an image. TypeSafe trains Jev for calibrated decisions rather than generated text, and ships Python and JavaScript SDKs, a browser Playground and an agent skill.
TypeSafe's model on Felo
Felo lists Jev 1.14 as a decision model, the newest Jev release. It answers choice, score, noul and box questions and is aimed at model-to-model routing, content classification and moderation, multilingual labeling and image-based decisions such as flagging screenshots. Image input and long context are experimental on the route and gated by request headers.
Why use TypeSafe through Felo
A Felo key covers Jev and the rest of the catalog, so adding a decision step does not mean another vendor account, contract or invoice. Usage lands on the same Felo balance and history, and switching between TypeSafe and another model is a model-ID change rather than a new integration.
Pricing and access
On Felo, Jev is billed per token from prepaid USD credits, with no subscription or monthly minimum. TypeSafe's own list price is per input token, with output tokens free on its API, and the published Felo rate currently works out to about $0.042 per 1M input tokens. The model table shows the live figure.
Who TypeSafe is for
Jev fits decisions a person could make in a few seconds: which queue, which label, whether an action may proceed. It is not a chat or generation model, it does not write replies, produce code or explain its reasoning, and calibration applies to groups of predictions rather than guaranteeing one answer. Image input and long-context mode are experimental, so production traffic belongs on text decisions with your own thresholds and escalation rules.
Specifications
- Input per 1M
- $0.042
- Output per 1M
- $0
- Context
- 262.144K
- Max output
- 235.929K
- Protocols
- —
Similar models from other providers
Frequently asked questions
- What is TypeSafe?
- TypeSafe AI is the company behind Jev, a System One model that returns typed decisions instead of prose. System One models are built for fast, structured judgments that software can consume directly.
- Which TypeSafe model can I call through Felo?
- Felo lists one TypeSafe model, Jev 1.14. The live model table on this page is the authoritative list and shows its current rate.
- Do I need a TypeSafe account or billing?
- No. Jev is called with a Felo API key and billed from your Felo balance. A separate TypeSafe account is only useful if you want its own SDKs, Playground or enterprise terms.
- How much does TypeSafe cost on Felo?
- Jev is billed per token from prepaid USD credits at the published rate, currently about $0.042 per 1M input tokens. On TypeSafe's own API output tokens are free. Check the model page for the exact figure.
- Can Jev write text or explain its answers?
- No. Jev does not generate replies, code or reasoning explanations. It returns choices, scores, probabilities and boxes that your code branches on, with confidence where the question type provides it.
- Is image input ready for production?
- Not yet. TypeSafe marks image input and long context as experimental and not suited to production, and they are enabled by request headers on the route. Text-only decisions are the supported path.