Page 04 · Access and cost snapshot
Where can I use Jev, and at what price?
Access routes are not interchangeable. The official model, a gateway, and an independent proxy can have different endpoints, billing, accounts, limits, and data boundaries.
Checked 20 September 2026: the current official model record names Jev 1.13, model ID
jev-1.13.0, at $0.042 per million input tokens with free output. The console has a login surface, while the public site still exposes a Join Waitlist path. That combination does not prove that every new account receives credits or API access.Official ID
jev-1.13.0
Current pinned model. jev-latest and jev-preview currently point to it.
Direct price
$0.042/M in
Output is listed as free. Actual spend still depends on state and question length.
Published limits
250k tok/s · 1,200 RPM
64k request context; 32k state plus longest question. TypeSafe warns limits can change dynamically.
Route by route
| Route | How to start | Endpoint / model | Price, limits, and input | Terms and confidence |
|---|---|---|---|---|
| TypeSafe direct official |
Console login offers Google or email-code sign-in. The public TypeSafe site also shows Join Waitlist. Login proves the access surface exists, not that a new account receives credits or API eligibility. | POST https://api.typesafe.ai/v1/systemonejev-1.13.0Aliases: jev-latest, jev-preview (currently the same target). |
$0.042 / 1M input tokens; output free. Published limits: 250,000 tokens/second and 1,200 requests/minute, subject to dynamic adjustment. 64k request context; 32k state plus longest question. Text-only input. | Official/current model and API pages. Read the legal index before sending data; access eligibility, credit grant, and live quotas remain an account-level check. |
| OpenRouter gateway |
Requires an OpenRouter account and API key. The Jev Lab documents a typed-decision workflow. | https://openrouter.ai/api/v1typesafe/jev-1.13 and Jev Latest. The Lab uses alpha.decisions. |
Current listing displays $0.042 / 1M input and $0 / 1M output; model page shows 32k context. Account quotas, routing fees/terms, and logging controls were not established here. | Third-party gateway listing, not direct TypeSafe billing. Verify current provider status, account limits, and data controls before production use. |
| Cloudflare AI gateway |
Requires a Cloudflare account/token. The public model page provides AI.run and REST examples. | typesafe/jev; the model page's example output identifies jev-1.13.0. Use Cloudflare's documented interface rather than assuming the direct TypeSafe endpoint. |
Model page shows a 32k context. Jev-specific numeric pricing is linked to the Cloudflare dashboard and is unknown in this public check; account limits are also unknown. | Third-party model page is current evidence of a route, not proof of the direct price or equivalent retention. Check Cloudflare and TypeSafe terms in the account context. |
| Vercel AI Gateway unverified |
The named page was inaccessible to the checker. Do not treat the URL as a confirmed way to obtain Jev. | Model ID, endpoint, provider status, and live availability were not accepted from the inaccessible page. | Price, limits, markup, and account conditions are unknown. No Vercel figure is carried forward. | Keep as an unverified lead until the page is reachable and exposes a current listing. This row is intentionally not a recommendation. |
| Independent hosted proxy not official |
Independent pricing page advertises its own account/key and prepaid packs, and claims a no-waitlist route. Those are the wrapper's claims. | https://jevtypesafeai.com/api/v1/decide; compatibility with Jev is the independent site's claim, not TypeSafe endorsement. |
$0.42 / 1M input tokens; output free; prepaid packs of $5–$100. That self-reported input price is 10× the current official displayed rate. | Separate account, billing, key, retention, and data-processing boundary. The accepted pricing page does not establish its full terms or live /get-jev route. |
Official context and data terms
“Not used to train” is not “not stored.” The checked TypeSafe legal pages say Input is not used to train or fine-tune models without consent. They also allow processing for service delivery, fee calculation, telemetry, fraud/abuse prevention, and legal purposes. The privacy policy says services are hosted in the United States and standard retention lasts as long as reasonably necessary; it does not give a Jev-specific retention duration. Enterprise zero-data-retention is a separate option.
- Master Customer Agreement: credits and usage limits govern access; service and telemetry processing are described; output can be inaccurate and service is not promised to be uninterrupted.
- Privacy policy: no training/fine-tuning on prompts or Input, service-provider disclosures, US hosting, and a general retention statement.
- Data processing addendum: processor terms, purpose-based retention, subprocessors, and international-transfer terms for the relevant customer arrangement.
Before you build against a route
- Check eligibility. Confirm that the account can obtain credits and an API key; a login page or waitlist link is not a quota promise.
- Pin the model when thresholds matter. Prefer
jev-1.13.0for a reproducible test; use an alias only when automatic upgrades are acceptable. - Measure the real state. Send text-only input, test the 32k state-plus-longest-question constraint, and treat the published rate limits as provisional.
- Recheck the data boundary. Gateways and the independent proxy add their own account, logging, retention, billing, and terms. Minimise state and keep a fallback for 429/529 or provider outages.
Cost intuition: input-token pricing is not a price per decision in the abstract. State and question text determine token use; gateway fees, quotas, and proxy packs can change the bill. The independent wrapper's self-reported $0.42/M is not official TypeSafe pricing.
Direct sources checked 20 Sep 2026
- TypeSafe models — current model ID, aliases, price, limits, context, text-only input, and data notes.
- TypeSafe API reference — endpoint, request/response shapes, and error conditions.
- TypeSafe console and TypeSafe homepage — login surface and the public waitlist path.
- TypeSafe legal index, MCA, privacy policy, and DPA — official data and contract terms.
- OpenRouter Jev 1.13, TypeSafe catalogue, and Jev Lab — current listing and typed-decision route.
- Cloudflare Jev model page and AI Gateway overview — model interface and account route.
- Vercel AI Gateway Jev page — named page inaccessible to the checker; no claim accepted.
- Independent wrapper pricing — self-reported price and non-affiliation statement.
- TypeSafe launch post — preserved bootstrap context for the original price and early-access framing.
This is an independent access refresh from run:6ee2dcc6-09c0-4f2a-8e92-6d6835767621, checked 20 September 2026. It does not upgrade the catalogue's separate finding that no checked public source establishes a verified commercial production product using Jev.