Jev pricing explained: cost per request and monthly examples.

The calculation

Under the pricing above, output tokens add nothing to that formula: there is no output-token charge to add. “Billable input tokens per request” means the whole request body Jev evaluates: the state you send plus every question's own instructions and criteria, not just the length of a customer message on its own.

Word counts do not convert exactly to token counts, so treat any estimate here as a planning number, not a bill. Once you are making real calls, every response includes a usage object with input_tokens and output_tokens for that exact call (see the quickstart). Use that measured number in place of an estimate as soon as you have it.

Cost by request size and volume

Illustrative token counts and volumes chosen to span a plausible range, not measured customer usage, a benchmark, or a throughput guarantee. Computed directly from the formula above.

Three monthly scenarios

Assumes a 30-day month. Volumes and token sizes are illustrative assumptions, not observed usage or an account capacity guarantee. Swap in your own numbers once you have measured usage.input_tokens from real traffic.

What changes your bill

• Context length. A longer state (message, document, history) means more billable input tokens for the same question.
• Question instructions and criteria. Each question's own wording and options are billed input too. TypeSafe's documentation says adding a question to an existing call “costs only the tokens for the extra questions, which are cheap,” and its own cookbook reports one specific case where batching 13 questions into one call was cheaper than 13 separate calls. That is a documented example, not a fixed multiplier you should expect on every workload.
• Request volume. Cost scales directly with how many requests you send.
• Retries. A retry that succeeds is billed like any other call. TypeSafe does not document whether a request rejected with 429 or 529 is billed, so check your own usage.
• Fallback calls to other models. If your app falls back to a different provider, that provider's own pricing applies on top of TypeSafe's.
• Third-party gateway or hosting charges. An intermediary between you and TypeSafe may add its own fee; see below.

All of this is TypeSafe's own inference cost. Your application's total cost usually also includes your own hosting, logging, any other AI provider calls, and engineering time. Keep the two separate when budgeting.

Direct access vs. third-party pricing

If you see a price for Jev or System One somewhere other than typesafe.ai, check what it actually is before trusting it: TypeSafe's own docs describe the direct API price above; a gateway or proxy that resells the TypeSafe API should separately document any markup or per-request fee it adds on top; and an unrelated service selling its own hosted feature or credits built on Jev has its own price, not TypeSafe's.

This guide has not verified pricing from any gateway or reseller, so it does not list or compare one here. Confirm any third-party number against that provider's own official page, and never assume the direct price above transfers automatically to a different integration path.

Practical ways to control cost

• Send only the context a question actually needs. Every extra token of irrelevant state is billed with no decision benefit.
• Keep instructions and criteria clear and avoid repeating the same wording across questions.
• Combine suitable independent questions over the same shared state into one call where it fits your case, rather than issuing several separate requests.
• Bound retries and handle rate limits deliberately, instead of looping on 429/529 responses.
• Measure actual usage with usage.input_tokens on real traffic before extrapolating a monthly number from assumptions.
• Include any fallback-model calls in your cost estimate, not just the primary Jev call.

Shorter input only helps if it still gives Jev everything it needs to decide correctly. Trimming away the one detail that would have changed the answer is not a saving.

Short FAQ

How much does Jev cost?

$0.042 per million input tokens for direct TypeSafe API access, with output tokens documented as free, as of docs.typesafe.ai/models on 2026-09-22. Your real cost depends on your own token volume; see the tables above.

Does Jev charge for output tokens?

No, not under the pricing verified above. A request's cost is driven entirely by its billable input tokens.

What would one million requests cost?

It depends on the input size per request: from $10.50 at 250 tokens/request up to $840.00 at 20,000 tokens/request, at the price above. See the first table for the full spread.

Does asking multiple questions change the cost?

Yes, modestly: each question's own instructions and criteria add their own tokens to the same call. TypeSafe documents that adding a question to an existing call costs only that question's own tokens, not the overhead of a separate request. See “What changes your bill” above.

Is Jev free to try?

This site's own recipes let you explore illustrative answers for sample messages at no cost, because showing them makes no API call. When this site's own live testing is enabled, Jev Guide pays for those calls. Running a recipe with your own API key bills your own TypeSafe account, per the pricing above. Whether TypeSafe offers free trial credit for new accounts is not something this guide could verify publicly; check the TypeSafe console for your own account's status.

Why might another website show a different price?

See “Direct access vs. third-party pricing” above. It may be an outdated snapshot, a gateway's own markup, or an unrelated product. Verify any number you see elsewhere against docs.typesafe.ai/models directly.

Sources and verification

• Models: price per input token, output-token policy, rate limits, and context limits. Primary source for every dollar figure above, verified 2026-09-22.
• Quickstart: the usage.input_tokens / usage.output_tokens response fields.
• Primitives: what counts as billable input, and how adding questions to one call is priced.

Jev Guide is an independent resource, not affiliated with or endorsed by TypeSafe AI. TypeSafe states that its rate limits can change without notice, and published prices can change too, so the figures above are last verified against the source cited, not a live or auto-updated price. Check docs.typesafe.ai/models yourself before budgeting. No taxes, minimum payments, subscriptions, regional pricing, or free trial credits are documented publicly as of this review, so none are assumed or estimated here.