Jev pricing, rate limits, models, and SDK versions
Compare Jev pricing, context, limits, and tested SDKs by channel. Full snapshot: 2026-09-30; TypeSafe direct API limits rechecked: 2026-10-23.
Jev does not have one channel-independent price or limit
TypeSafe direct API, Vercel AI Gateway, and Cloudflare Workers AI are separate access channels. Verify price, context, credentials, and quotas per host/provider. An unknown below means the reviewed public source did not publish the value—not zero and not unlimited.
Keep price and limits separated by channel
TypeSafe direct API
api.typesafe.ai- Model
jev-1.14.0alias jev-latest- Input price
- $0.042 per 1M input tokens ($42 per 1B)Published direct-API price
- Output price
- Free
- Context
- 64k tokens per request; state plus longest question limited to 32k
- Token rate limit
- Organization-specific; see the TypeSafe developer console
- Request rate limit
- Organization-specific; see the TypeSafe developer console
- Account quota
- unknown — not published in the public docs reviewed
Vercel AI Gateway
Vercel AI Gateway- Model
typesafe-ai/jev- Input price
- Starting at $0.04 per 1M input tokensGateway listing; provider table showed TypeSafe AI and DigitalOcean
- Output price
- unknown — not displayed on the reviewed model page
- Context
- 32K in the reviewed provider table
- Token rate limit
- unknown — not published on the reviewed model page
- Request rate limit
- unknown — not published on the reviewed model page
- Account quota
- unknown — depends on Gateway account and provider terms
Cloudflare Workers AI
env.AI binding / api.cloudflare.com- Model
typesafe/jev- Input price
- $0.042 per 1M input tokensCloudflare model-catalog token equivalent; Workers AI bills neurons in the back end
- Output price
- $0.00 per 1M output tokens
- Context
- 32,000 tokens in the reviewed model catalog
- Token rate limit
- unknown — no Jev-specific token rate published on the reviewed pages
- Request rate limit
- 300 requests per minute default for Text Generation unless a model requires Workers Paid
- Account quota
- 10,000 Neurons per day free allocation across Workers AI; paid overage requires Workers Paid
TypeSafe direct API limits were rechecked on 2026-10-23: request and token rates are organization-specific; consult your organization's current values in the developer console. The old snapshot's 80 requests and 100,000 tokens per second are not universal account entitlements. Exceeding either limit returns 429. Check your account before scheduling concurrent calls. Other channels retain the full snapshot date above; package observations are dated individually below.
Model aliases rechecked on 2026-10-26: jev-latest → jev-1.14.0 and jev-preview → jev-1.14.0. Both currently resolve to the same stable release; there is no separate preview build. Aliases move as releases ship. Pin a versioned model ID for reproducibility or threshold calibration, and log the response model. Pinning an SDK does not freeze a remote alias. Default context stays at 64k per request and 32k for state plus any one question. Experimental 262k context requires an opt-in header and comes with slower responses and reduced accuracy; it is not the default allowance. Official models and aliases ↗
Latest upstream is not the same as tested here
| Package | Tested here | Latest observed | Observed on | Tested-version runtime | Latest-observed runtime | Status | Published |
|---|---|---|---|---|---|---|---|
@typesafe-ai/sdk ↗npm | 0.6.0 | 0.7.0 | Node >=20 | Node >=22 | Newer upstream than tested | ||
typesafe-sdk ↗pypi | 0.7.2 | 0.7.4 | not separately recorded | Python >=3.10 | Newer upstream than tested | ||
langchain-typesafe ↗pypi | 0.0.1a3 | 0.0.1a3 | not separately recorded | Python >=3.10,<4; pre-release / beta | Tested = latest observed | ||
langchain-core ↗pypi | 1.6.6 | 1.6.10 | not separately recorded | Python >=3.10,<4 | Newer upstream than tested | ||
@ai-sdk/gateway ↗npm | 4.0.101 | 4.0.112 | not separately recorded | Node >=22 | Newer upstream than tested | ||
ai ↗npm | 7.0.123 | 7.0.139 | not separately recorded | Node >=22 | Newer upstream than tested | ||
@ai-sdk/typesafe-ai ↗npm | not tested | 3.0.18 | not tested | Node >=22 | Observed, not tested here |
This site never rewrites examples merely because a registry has a newer release. “Behind” requires a changelog review, lockfile update, and complete contract/browser rerun. “Not tested” records discovery, not compatibility or endorsement.
Changes to check before upgrading
Do not infer what the public pages do not say
- Gateway request rate, token rate, and account quota.
- The final bill under different accounts, contracts, regions, or promotions.
- Whether a newly published package works with these examples before a pinned test run.
- Whether price, limits, context, or aliases will remain unchanged.
Check upstream drift before shipping
The repository’s read-only monitor checks official-page markers, docs sitemap lastModified values, and npm/PyPI latest versions plus runtime constraints. It reports changes; it never edits prices or upgrades dependencies.
npm run seo:upstream - If the result is current, still confirm model access and billing terms in the real account.
- If it reports drift, read the primary change before updating the snapshot, page, examples, or lockfile.
- Any factual number change updates the verification date and reruns all gates; never refresh the date alone.