Choose Auto, TokenLab Verified, or Official for each request, with prices shown up front. See what's new

TypeSafe: Jev 1.14

Jev 1.14 is TypeSafe's preview System One decision model for making structured judgments from a supplied state. It accepts text, JSON objects and arrays and returns typed decisions rather than prose. Image input and Box questions are not provided yet.
Compare models
jev-1.14
AvailableTypeSafeDecisionsSyncPreview
Input / Output
$0.042 / $0.00
Context
32K
Released
Oct 14, 2026
Modalities
Decisions

About Jev 1.14

Jev 1.14 is the preview release of TypeSafe AI's Jev, a System One model that turns a supplied state into typed decisions instead of prose. Each question returns a Noul probability, a Choice option, or a Score level, and your code branches on the result. TypeSafe released it on October 14, 2026. It is newer than the stable Jev 1.13, for teams that want to test the next build.

Where it works well

  • Answers come back typed: a Noul probability, a Choice option with its probability distribution, or a Score level, so code never has to parse a sentence.
  • Choice and Score answers include a confidence value, so you can act on firm answers and send uncertain ones to a person or a reasoning model.
  • One request can carry many questions about the same state, which suits routing, triage, and gating steps that need several judgments at once.
  • It reads strings, JSON objects, and arrays, so a record or the current state of an application can be sent as it is.
  • As the newest Jev, it lets you check the next build against your own labeled cases while the stable Jev 1.13 stays in place.

When to choose another model

  • Jev 1.14 returns decisions only. It does not draft replies, explain why it chose an answer, or write code, so pair it with a chat model where a person needs prose.
  • It reads text only. Image input and Box questions are not provided yet, so convert other media into text or structured fields first.
  • TypeSafe's published list of known weak spots covers Jev 1.13 only, so check 1.14 against your own labeled cases before its answers trigger actions.

Getting started

  1. Create API key

    Create a key in Console, then use it with every model on the platform.

  2. Send your first request

    Copy the example for your language and run it against the endpoint.

    Decision evaluationTokenLab endpoint
    POST/v1/systemone
    curl -X POST "https://api.tokenlab.sh/v1/systemone" \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{
  "questions": {
    "refund": {
      "instructions": "Is a refund requested?",
      "type": "noul"
    }
  },
  "state": "A customer requests a refund.",
  "model": "jev-1.14"
}'

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Pricing

per 1M tokens
Verified price
—
Official price
Input$0.042per 1M tokensOutput$0.00per 1M tokens

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours
Requests
Success rate
P95 latency
Total tokens
Model performance
Metrics appear once privacy and data volume thresholds are met.

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open Jev 1.14 in Console with a prompt ready to edit or send.

Help me try jev-1.14 at /v1/systemone. Show the result and cost.

Use cases

  • Trial the next Jev build

    Replay a labeled sample of your tickets or records through jev-preview, compare answers and confidence with Jev 1.13, and decide whether the newer build is worth adopting.

  • Intent routing with confidence

    Classify each incoming message with a Choice question, then send it to fixed logic, a specialist model, or a person depending on how confident the answer is.

  • Guardrails around a chat model

    Ask Noul questions about a prompt or a draft reply, such as whether it requests something disallowed, then pass, review, or block the message on the probabilities.

  • Fuzzy rules inside a workflow

    Replace a brittle hand-written check with a question about the record, and branch in code on a Noul probability compared with a threshold you have tested.

Prompt examples

State: this support chat transcript. Choice question: which team should own it: billing, technical, or account?

State: an order record as JSON with a delivery date and a customer message. Noul question: does `customer_message` report a missing item?

State: a draft chatbot reply. Score question: how risky is this reply for a regulated industry, from 0 for harmless to 3 for must block?

Cost calculator

1M tokens
0.5M tokens
Estimated monthly cost$0.042

FAQ

What is Jev 1.14?

It is the Jev release that TypeSafe's jev-preview alias points to. Jev is a System One model: you send a state and typed questions, and it returns Noul, Choice, and Score answers with probabilities and no free text. TypeSafe released it on October 14, 2026.

How does Jev 1.14 differ from Jev 1.13?

Both take the same requests and the same three question types, so the difference is the build. Jev 1.14 is newer and still in preview, while Jev 1.13 is the stable release behind jev-latest. TypeSafe's documentation does not describe accuracy differences, so measure them on your own cases.

What do the jev-preview and jev-latest names mean?

jev-preview is a request alias for jev-1.14, so either name reaches this model. jev-latest is unchanged and still resolves to Jev 1.13. An alias can be repointed later, so send the exact ID jev-1.14 once you have tuned thresholds on this version.

Can Jev 1.14 read images?

No. It reads text: strings, JSON objects, and arrays. Image input and Box questions are not provided yet, and TypeSafe says audio and video are not accepted either. Convert other media into text or structured fields before you send it as the state.

When should I choose Jev 1.14 over Jev 1.13?

Choose 1.14 when you want to evaluate the newer build, or when your own tests show it fits your cases better. Stay on 1.13 when you want the stable release and thresholds you have already tuned, because TypeSafe's list of known weak spots covers 1.13 only.

How much does Jev 1.14 cost?

On TokenLab, Jev 1.14 costs Input $0.042 / Output $0.00 per 1M tokens. The pricing table above shows the full breakdown.

Which endpoint should Jev 1.14 use?

Use https://api.tokenlab.sh/v1/systemone for Jev 1.14. The request example below shows the matching code shape.

Which operations does Jev 1.14 support?

Jev 1.14 supports Decision evaluation. Select an operation above to see its endpoint and request example.

Compare Jev 1.14

Jev 1.14 vs Jev 1.13

Sources

Reviewed Oct 19, 2026

More from Jev

Jev 1.13
32K