# Jev Model Guide

> Independent developer notes for TypeSafe AI’s Jev System One model. Not affiliated with TypeSafe.

Jev is a hosted decision model. It returns a typed choice, score, or yes/no probability. It does not generate text.

## Core

- [What is Jev AI?](https://jevmodel.org/what-is-jev/): TypeSafe System One model, current public alias jev-latest
- [Jev 1.14](https://jevmodel.org/jev-1-14/): jev-1.14.0 (alias jev-preview, released 2026-10-14) — experimental PNG/JPEG image input, Box bounding-box questions and 262k long context, with a live playground; jev-latest stays on jev-1.13.0
- [Is Jev open source?](https://jevmodel.org/is-jev-open-source/): hosted API, no public weights
- [What is Laya?](https://jevmodel.org/what-is-laya/): Apache-2.0 local System One model from Convai Innovations. Open-source Jev alternative, not Jev. GitHub: https://github.com/NandhaKishorM/laya
- [Jev vs Laya](https://jevmodel.org/jev-vs-laya/): TypeSafe hosted Jev versus Apache-2.0 local Laya (github.com/NandhaKishorM/laya)
- [How to use Jev](https://jevmodel.org/how-to-use-jev/): Playground, Dashboard API key, endpoint, and request shape
- [Get Jev access](https://jevmodel.org/get-jev/): jevmodel.org Playground, Dashboard API key, and alternative gateways
- [Jev API is open](https://jevmodel.org/jev-api-open/): the earlier waitlist is closed; Console Playground and Create key are live
- [Jev API examples](https://jevmodel.org/api/): jevmodel.org endpoint plus choice, score, and noul request shapes
- [Jev API documentation](https://jevmodel.org/docs/): full reference — POST /v1/systemone, Bearer sk- keys, request body, typed answers, billing, limits, errors
- [Jev pricing](https://jevmodel.org/jev-pricing/): TypeSafe $0.042 / 1M input tokens, output free; or jevmodel.org packs from $9.90 with no vendor account
- [Jev on OpenRouter](https://jevmodel.org/jev-openrouter/): typesafe/jev-1.13, POST /api/alpha/decisions or /api/v1/systemone, 32K context
- [Jev on Vercel AI Gateway](https://jevmodel.org/jev-vercel-ai-gateway/): typesafe-ai/jev via experimental_evaluate in the AI SDK
- [Jev for agents](https://jevmodel.org/agent-tools/): remote MCP server at /mcp (jev_decide, jev_noul, jev_choice, jev_score) plus downloadable agent SKILL.md — same Bearer key and balance as /v1/systemone
- [Jev GitHub](https://jevmodel.org/jev-github/): github.com/typesafe-ai SDKs and skills; no model weights
- [OpenJEV](https://jevmodel.org/openjev/): independent open-weights model (CC BY-NC 4.0) with the Jev API shape; not Jev

## Compare

- [Jev alternatives](https://jevmodel.org/jev-alternatives/): every ranked System One alternative with scores, licences and how to choose
- [JevBench leaderboard explained](https://jevmodel.org/jevbench/): JevBench v1.4.2.2 — 95 systems, 91 ranked, 842 decisions each; Jev #4 composite, #1 intelligence in the top ten. Source: benchmarkheaven.com/jev-models
- [Jev vs Imajev-4B](https://jevmodel.org/jev-vs-imajev/): open Apache-2.0 Qwen3.5-4B at #1 on JevBench, reads images (github.com/mohit67890/imajev)
- [Jev vs decider-4b](https://jevmodel.org/jev-vs-decider-4b/): Apache-2.0 Qwen3.5-4B rebuild at #3 (github.com/Mapika/decider)
- [Jev vs JevK5](https://jevmodel.org/jev-vs-jevk5/): Apache-2.0 distilled 4B at #5, ties Jev on judge cases (github.com/allebee/jevk5)
- [Jev vs Cygnet](https://jevmodel.org/jev-vs-cygnet/): frozen Gemma 4 12B one-token readout at #6, no training (github.com/blockbrain-ai/cygnet-recipe)
- [Jev vs Winnow](https://jevmodel.org/jev-vs-winnow/): local 12B GGUF for decisions + chat + vision at #8 (github.com/EldanRing/winnow-inference)
- [Jev vs SemIf](https://jevmodel.org/jev-vs-semif/): MIT logit reader at #13, edges Jev on judge cases (github.com/TheoLeeCJ/SemIf)
- [Jev vs Quyet-1.0-Large](https://jevmodel.org/jev-vs-quyet/): Apache-2.0 Gemma-4-31B, JevBench v1.6.1 open-weights Capability #1 at 81.7 (huggingface.co/chinhnc/Quyet-1.0-Large)
- [Jev vs deck-31B](https://jevmodel.org/jev-vs-deck-31b/): frozen Gemma-4-31B-it, Capability #2 at 77.6, weaker calibration than Jev
- [Jev vs Perplexity Decider v1.1](https://jevmodel.org/jev-vs-perplexity-decider/): Qwen3.8-27B, Capability 82.8, outside the v1.6.1 cost cap (huggingface.co/perplexity-ai/pplx-decider-v1-27b)
- [Jev vs Kev 27B](https://jevmodel.org/jev-vs-kev/): Apache-2.0 pointer head, Capability 78.9, estimated cost far above Jev (huggingface.co/jaredpalmer/kev-27b)
- [Jev vs Clef](https://jevmodel.org/jev-vs-clef/): Cloudflare Clef-Flash Capability #15 and the unranked 27B Clef (huggingface.co/Cloudflare/clef)
- [Jev vs Xor 26B-A4B](https://jevmodel.org/jev-vs-xor/): Juspay Gemma-4-26B-A4B, Capability #4 at 72.4. Xor 1.2 is a different checkpoint
- [Jev vs Jev-Omni](https://jevmodel.org/jev-vs-jev-omni/): independent Gemma-4-12B for text, image, audio, and video, Capability #6 (huggingface.co/akhilaaa3/Jev-Omni)
- [Jev vs d1-3B](https://jevmodel.org/jev-vs-d1/): Liquid AI open 3B decision model, Decision Index v0.2.1 score 48.57, not a JevBench rank (huggingface.co/LiquidAI/d1-3B)
- [Jev vs Plumb-4B](https://jevmodel.org/jev-vs-plumb/): September composite #2, October Capability #12
- [What is Laya?](https://jevmodel.org/what-is-laya/)
- [Jev vs Laya](https://jevmodel.org/jev-vs-laya/)
- [Jev vs GPT](https://jevmodel.org/jev-vs-gpt/)
- [Jev vs Claude](https://jevmodel.org/jev-vs-claude/)
- [OpenAI Decisions API vs Jev](https://jevmodel.org/openai-decisions-api/): OpenAI DevDay 2026 decision endpoint (Luna, text + image context, limited preview, pricing unannounced) compared with Jev

## Use cases

- [Use cases index](https://jevmodel.org/use-cases/): four categories of typed-decision patterns
- Evaluation & quality: https://jevmodel.org/use-cases/evaluation/
  - [LLM as a judge](https://jevmodel.org/use-cases/llm-as-a-judge/)
  - [AI agent evaluation](https://jevmodel.org/use-cases/ai-agent-evaluation/)
- Data matching: https://jevmodel.org/use-cases/data-matching/
  - [Entity matching](https://jevmodel.org/use-cases/entity-matching/)
- Retrieval & knowledge: https://jevmodel.org/use-cases/retrieval/
  - [RAG evaluation](https://jevmodel.org/use-cases/rag-evaluation/)
  - [Live web context](https://jevmodel.org/use-cases/web-context/)
  - [Context compaction](https://jevmodel.org/use-cases/context-compaction/)
- Routing & decisions: https://jevmodel.org/use-cases/routing/
  - [AI model routing](https://jevmodel.org/use-cases/ai-routing/)
  - [Support triage](https://jevmodel.org/use-cases/support-triage/)
  - [Tool-call guardrails](https://jevmodel.org/use-cases/tool-call-guardrails/)
  - [Content moderation](https://jevmodel.org/use-cases/content-moderation/)
  - [Intent classification](https://jevmodel.org/use-cases/intent-classification/)
- Review scoring (evaluation): https://jevmodel.org/use-cases/review-scoring/

## Account and billing

- [Jev sign up](https://jevmodel.org/jev-sign-up/): create a jevmodel.org account with Google or email, no invite code or waitlist; what to do if the verification email does not arrive
- [Billing and refunds](https://jevmodel.org/refund-policy/): input-token billing, automatic return of failed requests, pack and plan terms, how to request a refund
- [Terms of Service](https://jevmodel.org/terms/)

Official site: https://typesafe.ai
jevmodel.org API: POST https://jevmodel.org/v1/systemone (API key from /dashboard)
Official TypeSafe API: POST https://api.typesafe.ai/v1/systemone

## Try Jev
- Jev playground (live API, 10,000 free input tokens, 100,000 after sign-in): https://jevmodel.org/playground/
- Jev batch (upload CSV/TXT/JSONL, run the same typed questions on every row, export CSV/JSONL): https://jevmodel.org/batch/
- Playground & API tokens (Starter pack $9.90 for 15M input tokens, never expires): https://jevmodel.org/pricing/
- Drop-in endpoint with a jevmodel.org key: POST https://jevmodel.org/v1/systemone (same body as TypeSafe SystemOne)

## Other languages (core pages: home, what-is-jev, how-to-use-jev, api, jev-pricing, playground, batch, pricing, dashboard, docs, agent-tools, jev-1-14)
- 简体中文: https://jevmodel.org/zh/
- 繁體中文: https://jevmodel.org/zh-tw/
- 日本語: https://jevmodel.org/ja/
- 한국어: https://jevmodel.org/ko/
- Deutsch: https://jevmodel.org/de/
- Español: https://jevmodel.org/es/
- Français: https://jevmodel.org/fr/
- Português: https://jevmodel.org/pt/
- Italiano: https://jevmodel.org/it/
- Polski: https://jevmodel.org/pl/