Jev 1.13

Jev 1.13 is the versioned TypeSafe System One contract. Current docs still map jev-latest to 1.13, while jev-preview points to 1.14; compare the versioned 1.14 record before changing aliases.

Jev 1.14
New version, default and opt-in limits.
Jev version guide
Compare Jev 1.13 with the other versioned Jev releases and check moving versus pinned aliases.

Decision ModelText ClassificationStructured OutputAgent RoutingLow Latency

TypeSafe hosted API

$0.042 per 1M input tokens; output tokens free

Commercial

🚀Function Overview

A text-only decision model for ranking, routing, filtering, scoring and other bounded choices where compact calibrated probabilities are more useful than generated prose.

Key Features

Pinned historical model ID jev-1.13.0; the latest alias still selects this version.
Moving alias jev-latest currently points to Jev 1.13
POST /v1/systemone request surface
Typed Choice, Score and Noul decision primitives
Text, strings, JSON-like state and arrays as input; no image, audio or video input
Probability and typed decision output; no prose, code or reasoning explanation generation
64,000-token overall request limit documented by TypeSafe
Default 32K per-question budget within the overall request; see the current model documentation for counting.
$0.042 per million input tokens with no output-token charge
English is the primary language; CJK support is documented as uneven
Customer requests and responses are not used for training; zero-data-retention is an enterprise option
Group-level calibration requires task-specific thresholds and fallbacks
Current rate limits are organization-specific in the TypeSafe console; the earlier fixed-rate statement is historical.
October 21 documentation: default 64K/request, 32K/question; the explicit long-context header raises both to 262K. Current rate limits are organization-specific; earlier fixed rates are historical.

Use Cases

•Rank candidate actions, documents, code paths or search results
•Route a request to a model, tool, workflow or human reviewer
•Filter browser content and agent tool output
•Score policy, relevance, quality or safety criteria
•Choose navigation edges in browser, desktop and graph agents
•Triage cases before a slower reasoning or generative model

⚙️Input Parameters

model

string

Use jev-1.13.0 to pin the dated model or jev-latest for TypeSafe's moving alias.

state

string | object | array

Textual context, strings, JSON-like objects or arrays that describe the item or environment being judged.

questions

object

Typed Choice, Score or Noul questions and their allowed labels, options or scale.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "jev-1.13.0",
  "state": "A public document URL is available and the user asks for a summary.",
  "questions": {
    "read_only_fetch": {
      "type": "noul",
      "instructions": "Does the requested task require only reading the supplied public document?"
    }
  }
}

Output Results

Illustrative documented request shape; not executed here. Inspect the actual typed probability response and model version. The application must enforce permissions and validate its threshold.

Quick Actions

Technical Specifications

Hardware Type: TypeSafe hosted API
Commercial Use: Supported
Pricing: $0.042 per 1M input tokens; output tokens free