Last updated: 2026-11-09
With 1,000-row batch uploads, visual Decision Flows and one API key, JevModel is an independent console for TypeSafe AI's jev-1.13 model, which returns typed Choice, Score and Noul answers with probabilities instead of generated text. Requests are routed through OpenRouter, and JevModel states that it is not affiliated with TypeSafe AI.
About JevModel
JevModel is an independent web console and API for Jev, the decision model from TypeSafe AI. Instead of generating text, the model answers bounded questions about a piece of text or JSON: Choice picks one of your options, Score places the input on a scale you define, and Noul gives a yes/no probability for a criterion. Every answer comes back with probabilities your code can check before it acts.
The service wraps that model in four tools that share one token balance: a Playground for single runs, a canvas for Decision Flows that chain questions and bill only the branches an item reaches, a batch runner that accepts CSV, TXT or JSON Lines files of up to 1,000 rows, and a REST API at /api/v1/systemone. Requests are routed through OpenRouter to typesafe/jev-1.13, and the site describes itself as a separate service from TypeSafe AI.
Typical jobs are support triage, ticket routing, spam and toxicity screening, priority and escalation, language detection, personal-data screening and prompt-injection checks. It suits a developer who would otherwise write a schema-constrained prompt with BAML or train a small classifier with OpenPipe and wants to test the idea on a CSV first. Because the output is typed data, it can drive branching in a workflow tool such as n8n.
Billing is in US dollars through prepaid token packs, with monthly and yearly plans as the alternative; the amounts are in the price notes below. The privacy policy says saved runs are kept for 30 days, with the latest 100 shown, and advises against submitting passwords, card details or sensitive personal records.
Pricing
Free: no account is needed for the 3 daily demo runs (UTC day), plus 100,000 tokens once after email verification (about 500 short requests), 1 saved Decision Flow and 50-row batches. One-time packs that never expire: Icebreaker $4.90 for 8M tokens, Starter $9.90 for 20M, Builder $19.90 for 50M, Pro $49 for 180M and Scale $99 for 450M, which works out to $0.613 down to $0.220 per 1M input tokens. Monthly plans are $9.90 (Starter), $19.90 (Builder), $39.90 (Pro) and $79.90 (Scale), unused tokens roll over, and yearly billing saves 50%.
Output tokens are free and failed requests are refunded. The underlying Jev model is billed by its provider at a separate rate.
| Tier | Price | What it includes |
|---|---|---|
| Free | Free | Three trial runs a day, no account needed, A one-time grant after email verification, A single saved Decision Flow, Small batch uploads only |
| Monthly plans | $9.90/mo | Starter $9.90, Builder $19.90, Pro $39.90, Scale $79.90 per month, Unused tokens roll over month to month, Yearly billing saves 50% and delivers all 12 months of tokens up front |
| Icebreaker pack | $4.90 one-time | Covers about 40,000 short requests, Tokens never expire |
| Starter pack | $9.90 one-time | Covers about 100,000 short requests, Unlocks unlimited Flows and larger batches |
| Builder pack | $19.90 one-time | Covers about 250,000 short requests |
| Pro pack | $49 one-time | Covers about 900,000 short requests |
| Scale pack | $99 one-time | Covers about 2,250,000 short requests |
Key Features
- Choice, Score and Noul answers: Three typed answer formats return one of your options, a level on your own scale, or a yes/no probability, each with the probability of every answer.
- Decision Flows: Questions are chained on a canvas and run with one request, and only the question nodes on the path an item takes are billed.
- Batch runs: CSV, TXT or JSON Lines uploads of up to 1,000 rows (50 on a free account) return every structured answer, with a CSV export.
- One API and one balance: POST /api/v1/systemone with a Bearer key shares the same token balance as the Playground, batch runner and Flows.
- No-signup trial: The homepage demo allows 3 live runs per UTC day without an account, and an email-verified account gets 100,000 tokens once.
- Refunds on failed requests: A request that fails before a result is saved is refunded automatically, and output tokens are never charged.
Pros
- Returns the probability of every answer, so borderline cases can be sent to a person instead of acted on blindly.
- Packs do not expire and failed requests are refunded, which keeps early experiments low-risk.
- The Flow canvas and batch runner let you test a decision on your own CSV before writing any integration code.
- One balance across console and API means no separate billing setup with the model provider.
Cons
- The per-token rate is a multiple of the model provider's direct rate, which the pricing page itself acknowledges as the cost of the hosted tools.
- A free account is capped at a single saved Decision Flow and small batch uploads until any purchase is made.
- It is independent of TypeSafe AI and sends runs through OpenRouter, so the privacy policy tells users to keep passwords, card details and sensitive personal records out.
- No SOC 2, HIPAA or ISO attestations are listed, and batch results are row-level answers and CSV exports rather than an accuracy report.
Product Information
- Cloud
- Yes
- Self-Hosted
- No
- On-Premise
- No
Data Handling
- Training-data policy
- The privacy policy does not state a training policy. Runs are sent to OpenRouter, which routes them to the Jev model provider.
- Data retention
- 30 days
Frequently Asked Questions
What does JevModel actually cost?
The free account gets 100,000 tokens once. Five one-time packs never expire and step up from Icebreaker at $4.90 (8M tokens) through Starter at $9.90 and Builder at $19.90 to Pro at $49 and Scale at $99 (450M tokens). Monthly plans run $9.90 to $79.90 with unused tokens rolling over, and yearly billing saves 50%. Only input tokens are charged, output is free, and each request is billed at the usage the model reports.
Can you use JevModel without paying?
Yes. The homepage demo runs 3 times per UTC day with no account, and verifying an email adds 100,000 tokens once, about 500 short requests. A free account is limited to 1 saved Decision Flow and batch uploads of 50 rows; any purchase lifts both limits for good.
What are JevModel's closest competitors?
BAML fits teams that want typed outputs from any language model inside their own code. OpenPipe suits teams that would rather fine-tune a small classifier on their own labelled data. OpenRouter is the route if you want to call the same Jev model at the provider's rate without the console.
How does JevModel compare to OpenRouter in 2026?
OpenRouter is the gateway JevModel itself uses to reach typesafe/jev-1.13, and JevModel's pricing page admits a direct account can have a lower per-token rate. JevModel costs more per token but adds the Flow canvas, batch evaluation, refunds on failed requests and one balance across console and API. Choose OpenRouter for high raw volume from code, and JevModel when you want to design and test decisions before writing an integration.
How do you set up JevModel?
Run the three free demo runs on the homepage, then sign up and verify your email to receive the 100,000 free tokens. Open the Console Playground, pick a preset and replace its sample state with your own text. For code, create a key at /console/keys and POST a state and named questions to /api/v1/systemone with a Bearer header.
Top Alternatives
- OpenRouter: Pick OpenRouter if you want one gateway to many models, Jev among them, billed at each provider's own pricing.
- BAML: Pick BAML if you want typed, schema-checked outputs from any language model inside your own code.
- OpenPipe: Pick OpenPipe if you want a small classifier fine-tuned on your own labelled examples.