Jev model guide · updated September 28, 2026
Jev Model by TypeSafe AI
The Jev model from TypeSafe AI turns text into decisions your software can use. Give this System One model a state and focused questions; get yes/no probabilities, choices and scores. Use the Jev model to route a ticket, check a rule or rate a response, with uncertainty visible in the result.
Jev, TypeSafe AI and this playground
TypeSafe AI builds the Jev model. Jev AI provides the independent workspace you are visiting.
Looking for Jev TypeSafe resources? The TypeSafe AI website introduces the developer and its research. The official documentation describes the model and TypeSafe service.
Here you can try the Jev model, save judges and connect applications through our API. Your account and purchases are managed by Jev AI. Start with the TypeSafe AI Jev walkthrough to choose the right service and run your first example.
A Jev model example: one ticket, three decisions
In our saved support-triage example, a customer reports repeated payout failures and threatens to switch providers. The Jev model evaluates routing, urgency and frustration in one request.
| Question | Recorded output | How an app can use it |
|---|---|---|
| Which team should handle this? | billing | Assign the ticket to the billing queue. |
| Does the customer need a response today? | 96% probability of yes | Compare the result with your urgency threshold. |
| How frustrated is the customer? | 3 on a 0–3 scale | Interpret the score using the question's own rubric. |
Recorded 2026-09-18 with jev-1.13.0. These are model predictions, not measured accuracy rates. Viewing this example makes no live call. Inspect the questions and full answers, or run the Jev model on your own text using your available balance.
How the Jev model works
The name comes from Daniel Kahneman's Thinking, Fast and Slow. System 1 thinking is fast and intuitive; System 2 is slow and deliberate. Jev targets the first kind: quick, focused judgments that software can branch on.
A request has three parts. The state is the text or JSON Jev reads. The model is jev-latest, which currently resolves to jev-1.13.0. The questions are a map of typed primitives:
- Noul — a yes/no question. Jev returns the probability of yes, from 0 to 1.
- Choice — pick one option from a set you define. Jev returns the winner, every option's probability and a confidence.
- Score — rate against ordered levels you write. Jev returns a probability-weighted score that can land between levels.
The Jev model reads the state once and evaluates the questions in parallel. Keep each question focused: choose a team, check urgency or score frustration rather than asking for a general essay.
Jev model training: RLCD
TypeSafe describes three ways to post-train a pretrained language model.
- RLHF (reinforcement learning from human feedback) trains chatbots to give answers people prefer. TypeSafe points out that it can reward confident-sounding mistakes and narrow the range of outputs.
- RLVR (reinforcement learning with verifiable rewards) produced reasoning models. They are strong at maths but slower and more expensive.
- RLCD (reinforcement learning for calibrated decisions) is Jev's path. The model returns decisions and probabilities, and higher probability should mean a higher chance of being right.
Calibration is a property of many predictions, not a promise about one. Across many answers where Jev says 0.8, about 80% should be correct. That is what makes the confidence field useful: act automatically when it is high, ask for review when it is middling, and route to a person when it is low.
Jev model facts
| Current version | jev-1.13.0 (aliases jev-latest and jev-preview) |
| Input | Text only: a string, a JSON object or an array. No images, audio or video. See the multimodal options. |
| Context | 64k tokens per request; 32k for the state plus the longest single question |
| Official price | $0.042 per million input tokens on TypeSafe; output tokens are free |
| Official rate limits | 250,000 tokens per second / 1,200 requests per minute, adjusted as capacity changes |
| Language | Best in English; other languages work with lower accuracy |
| Customer data | TypeSafe states that Jev is not trained on customer requests and is not fine-tuned per account |
Facts from TypeSafe's public documentation as of September 18, 2026. Jev AI pricing is on the pricing page; see what you can try for free. Jev’s weights are not public: open alternatives you can run locally.
Jev model capabilities and limits
TypeSafe publishes a candid list of Jev 1.13's weak spots. Designing around them is most of the craft.
- Good at: common-sense judgments about text: routing, moderation, intent, sentiment, relevance, policy checks, citation support and guardrails.
- Literal reading: Jev answers the question you wrote. Put boundary cases in the criteria.
- Numbers, counting and dates: keep arithmetic and date comparison in code. Use Jev to extract or classify the pieces.
- Long, noisy state: accuracy falls when the state is full of unrelated text. Filter first.
- Adversarial text: injected instructions inside the state can move answers. Test edge cases before launch.
- Generation: Jev does not write. Pair it with a generative model when you need text.
Jev model vs a generative LLM
Choose the output your application actually needs.
Use the Jev model when a workflow needs a label, a probability or a score. Use a generative LLM when it needs a customer reply, a summary or code. A support workflow can combine both: Jev selects the queue and flags urgency; a separate model drafts the reply.
The Jev model can still make a wrong judgment. Typed output makes a response easier to consume; it does not make every decision correct. Evaluate uncertain cases and use human review where mistakes are costly. Explore decision use cases and Jev model comparisons for further examples, or read the full Jev vs LLM comparison.
Try the Jev model, then connect your app
- 01
Try it without code
Open the Jev AI playground. Load a demo, edit the questions and run it on your own text.
- 02
Copy the request
“Get API code” turns your playground setup into cURL, JavaScript or Python.
- 03
Call the Jev API
Create a key on the Jev API page (key setup guide) and send the same JSON to
/api/v1/systemone. You can also use TypeSafe's own API and SDKs directly.
Jev model FAQ
What is Jev?
The Jev model is a System One model from TypeSafe AI. It reads the state you provide and returns typed decisions: yes/no probabilities, choices and scores. The Jev AI model playground on this site lets you try those decisions with your own text.
Who develops the Jev model?
TypeSafe AI develops the Jev model. Jev AI is an independently operated playground and API service. Accounts, API keys, balances and purchases on this site belong to Jev AI, not to TypeSafe.
Is the Jev model an LLM?
The Jev model understands natural-language input but does not write chat replies. For a workflow that needs both a decision and a written response, use Jev to select or score, then let a generative LLM write the response.
What does RLCD mean?
RLCD stands for reinforcement learning for calibrated decisions, TypeSafe's post-training method. It rewards correct decisions with honest probabilities rather than answers people prefer to read.
How should I evaluate the Jev model?
Test the Jev model on examples from your own workflow, including ambiguous inputs and known failures. Check decision quality, confidence, latency and actual billed usage together. A benchmark result is not a guarantee for a different task.
Does the Jev model use the chat-completions format?
The Jev model uses state and typed questions rather than a chat conversation. This site accepts that request at /api/v1/systemone. If you choose another provider, check its dedicated decision endpoint and model support instead of sending the request to a generic chat-completions endpoint.