Pricing
Plans and pricing
What each plan costs, what it opens, and the real price of every model.
Money is counted in integer millimes end to end. 1 TND = 1,000 millimes, and no floating-point number ever touches your balance.
Plans#
| Plan | Price | Credit | Req./min | Models |
|---|---|---|---|---|
| free | Free | 4.743 TND | 60 | Flash only |
| t100 | 120.000 TND | 100.000 TND | 60 | Whole catalogue |
| t200 | 239.000 TND | 200.000 TND | 60 | Whole catalogue |
| t500 | 596.000 TND | 500.000 TND | 120 | Whole catalogue |
| t1000 | 1 191.000 TND | 1 000.000 TND | 120 | Whole catalogue |
Any paid plan opens the whole catalogue. No model is reserved for the most expensive plan: what you buy is credit and a rate limit, not access rights.
Models#
Rates below are millimes per million tokens, rounded to the integer.
| Model | Input | Output | Min. plan | Vision |
|---|---|---|---|---|
| deepseek-v4-flash | 1 581 mi | 6 324 mi | free | — |
| deepseek-v4.1-flash | 1 581 mi | 6 324 mi | t100 | — |
| kimi-k2.7-code | 5 007 mi | 21 080 mi | t100 | ✓ |
| kimi-k3 | 15 810 mi | 79 050 mi | t100 | — |
| glm-5.2 | 7 378 mi | 23 188 mi | t100 | — |
| glm-5.3 | 7 378 mi | 23 188 mi | t100 | — |
| deepseek-v4-pro | 12 648 mi | 25 296 mi | t100 | — |
| qwen3.8-max | 10 540 mi | 31 620 mi | t100 | — |
| qwen3.7-max | 13 175 mi | 39 525 mi | t100 | — |
| qwen3.7-plus | from 2 635 mi | from 10 540 mi | t100 | ✓ |
| qwen3.6-flash | from 1 318 mi | from 7 905 mi | t100 | ✓ |
| qwen3.7-flash | from 158 mi | from 685 mi | t100 | — |
| qwen3.8-flash | 791 mi | 2 477 mi | t100 | — |
| qwen3-coder-flash | from 1 581 mi | from 7 905 mi | t100 | — |
| glm-5.2-fast-preview | 14 756 mi | 46 376 mi | t100 | — |
Prices by request size#
Some models are priced by the size of the prompt: Model Studio bills them in bands (for example up to 32K, up to 256K and up to 1M input tokens). The table shows the entry band. A request is billed entirely at the band its prompt falls in — input, cached input and output alike — exactly as the provider bills us, so a short request on these models costs the entry rate and a very long one costs the higher rate on every token.
Credit expires#
Purchased credit stays spendable for six months from the moment it is granted, including the credit given at signup. One rule, applied to every millime in the wallet.
Spend cap#
POST /v1/spend-cap sets a monthly ceiling in millimes, or clears it with
null. Past it, requests get a 402 with the code spend_cap_exceeded — the
wallet is not empty; you asked for the brakes.