What Jev costs
Input tokens are billed, output tokens are free. Work out what your own workload costs before you build on it.
01The rate
Input tokens are billed at $0.042 per million. Output tokens are free. Both TypeSafe and Vercel publish the same figure, and there is no separate per-request charge, so your bill is a function of how much state you send.
02Work out your own number
The figure that matters is tokens per request, and it is dominated by your state rather than your questions. A short message with three or four questions attached lands somewhere near 350 tokens.
03What that means in practice
| Requests per month | Input tokens | Cost |
|---|---|---|
| 10,000 | 3.5M | $0.15 |
| 100,000 | 35M | $1.47 |
| 1,000,000 | 350M | $14.70 |
| 10,000,000 | 3.5B | $147.00 |
04Two ways to spend less
- Ask every question you need in one request. The state is billed once however many questions ride along with it, so four separate calls cost roughly four times what one call with four questions costs.
- Trim the state, not the questions. A conversation truncated to the last few turns usually judges the same and bills a fraction as much.
05What is not in the rate
Whatever your route charges to sit in front of the model. Going through a gateway or through OpenRouter adds that provider’s own fees and rate limits on top of the token rate above, and those change independently of what TypeSafe charges.