How much does Jev cost?

TypeSafe's published price is $0.042 per million input tokens. Output tokens are free, because Jev returns typed answers rather than generated text. There is no free tier.

Why are output tokens free on Jev?

Jev does not decode text token by token. It answers a fixed set of typed questions in one parallel pass, so there is no generated output to meter. Only what you send in is billed.

What limits how fast I can classify a large batch?

Usually the request cap rather than the token cap. TypeSafe publishes limits of 1,200 requests per minute and 250,000 tokens per second. On short items like social posts you run out of requests long before you run out of token throughput, so batching several items per request is what speeds a large job up.

How accurate is this calculator?

It is an estimate. It converts characters to tokens at roughly four characters per token, which is a reasonable rule of thumb for English prose and wrong for code, other languages and heavy emoji use. Every assumption is printed on the page so you can adjust it.

Jev cost calculator

TypeSafe's published price for Jev is $0.042 per million input tokens, with output free. Enter your workload to see what it costs, what the same job costs on a language model, and how long the rate limits say it will take.

Everything runs in your browser. No request leaves the page.

Items to judgePosts, replies, tickets, messages.Average characters per itemA typical X post runs about 220.Questions per itemAll of them answer in one request.Items per requestBatching amortises the rubric and dodges the request cap.Compare against

Jev

$0.9870

$0.000010 per item

GPT-5.6 Terra

$61.40

62x more than Jev

Wall clock at TypeSafe's published caps

83.3 minutes

Bound by requests

Relative cost

Jev$0.9870

GPT-5.6 Terra$61.40

Timing is arithmetic on the rate limits TypeSafe publishes, not a measured result. Comparison prices are each vendor's published list price as of September 2026.

The working

Every assumption is on this page so you can check the number rather than trust it.

Step

Value

Assumption

Tokens per item

55.0

4 characters per token. Rough for English prose, wrong for code and emoji.

Rubric tokens per request

180

About 60 tokens per question. A rubric with worked examples runs higher.

Requests

100,000

Items divided by batch size, rounded up.

Billed input tokens

23,500,000

Items plus one copy of the rubric per request. Output is free.

Time if request bound

83.3 minutes

1,200 requests per minute.

Time if token bound

1.6 minutes

250,000 tokens per second.

Price and caps from TypeSafe's published models page, September 2026. The timings above are our arithmetic on those caps, not a measurement.

The thing worth noticing: on short items the job is almost always bound by the request cap, not the token cap. You can be using a rounding error of your token throughput and still be waiting, which is why batching several items into one request speeds a large run up far more than anything else you can change. The limits are listed in full on the Jev limits page.

The language model comparison assumes the same text and rubric go in, plus a short typed answer coming back out. That output is the asymmetry people miss: on Jev it is free, because Jev never generates text. Prices for the comparison models move, so treat that column as an order of magnitude rather than a quote.

Keep exploring

Limits, worked examples and a real production integration.

Jev rate limits and context window

The caps this calculator uses, and what they mean in practice.

Classifying a million posts

The full worked cost model at volume.

Choice, Score and Noul

Pick the right primitive before you price the job.

Building a post scorer on Jev

Seven questions on one draft, rubrics and thresholds included.

Jev for social media

The whole cluster: what Jev is, and where it fits.

Decide with Jev, publish with OpenTweet

Once the judging is done, something has to post the result. That part is one API call to X, Bluesky and LinkedIn.

7-day free trial. Cancel anytime.