Jev AI guides for the decision layer.
Start with our Playground and Dashboard, then use the integration notes and evaluation checklists when you move Jev into production.
What is Jev AI? TypeSafe’s System One model explained
Typed decisions versus text generation, current model jev-1.13.0, and where Jev fits.
How to use Jev: Playground, API key, and first request
Try Jev here, create a jevmodel.org key, and copy the working Python or curl request.
Create and manage your Jev API keys
Sign in for 100,000 free input tokens, create separate keys, and copy the production endpoint.
Jev Playground: validate a decision before coding
Test choice, score, and noul questions with the same shape used by the API.
Alternative Jev access paths
Official TypeSafe access and Vercel AI Gateway for teams that need another billing path.
Jev for agents: remote MCP server and agent skill
Four typed tools over Streamable HTTP plus a SKILL.md playbook — same key and balance as /v1/systemone.
Jev on OpenRouter: typesafe/jev-1.13
Model ID, the Decisions and System One endpoints, 32K context, and a working request.
Jev on Vercel AI Gateway: typesafe-ai/jev
Call Jev with experimental_evaluate from the AI SDK. Pricing, limits, and code.
Jev GitHub: official TypeSafe repos and SDKs
What github.com/typesafe-ai publishes, what it does not, and which repos are look-alikes.
OpenJEV: the open-weights Jev alternative
Independent CC BY-NC model with the Jev API shape. 84.0% vs hosted Jev 85.4% on its own eval.
Jev API is open: waitlist dropped 21 September 2026
Waitlist URL is gone. Console Playground and Create key are live. $5 monthly credit, $0.042 / 1M input.
Is Jev open source? What developers should verify
Hosted model, no public weights. Laya is the open-weight alternative, not Jev.
What is Laya? Open-source local System One model
Apache-2.0 decision model from Convai Innovations. Same choice/score/noul job as Jev, not a TypeSafe release.
Jev vs Laya: hosted TypeSafe model vs open-source local decisions
Laya is an Apache-2.0 System One model on GitHub. Same choice/score/noul job, different product.
Jev vs GPT: a practical architecture decision
Use a decision layer around a general model instead of one model for every branch.
Jev vs Claude: gate the expensive reasoning pass
Let Jev approve or route, then spend Claude tokens only when the work needs language.
OpenAI Decisions API vs Jev: what DevDay 2026 changed
OpenAI’s Luna-based Decisions API takes text or images and picks from fixed answers. Limited preview, no pricing yet.
Jev pricing: $0.042 per million input tokens
Input-only billing, free output, and a simple token calculator.
How to benchmark a Jev decision layer
Accuracy, calibration, latency, fallback, and downstream outcomes.