DEVSAIO uses prepaid billing. You add funds, and each request costs exactly what it used. Here is what happens behind the scenes.
1. Before the request: a hold
When a request arrives, a small amount is reserved from your balance, based on the model and the maximum output allowed. This guarantees we never serve a request you can't pay for. You'll see it as “held for running requests” on the Billing page.
2. After the request: the real cost
When the model finishes, we count the actual tokens (or images, or seconds of video) and charge exactly that. The rest of the hold is released immediately.
3. If something fails
If the model provider refuses the request or it errors before producing anything, the hold is released and you're not charged.
How prices are measured
| Kind of model | Billed per |
|---|---|
| Text (chat) models | 1M input tokens and 1M output tokens, separately |
| Image models | image |
| Video models | second of video |
| Embeddings | 1M input tokens |
For example, the current price of claude-sonnet-4-6 is $3.15 / $15.75 per 1M tokens (input / output). Output tokens usually cost more than input tokens, so long answers matter more than long prompts.
Seeing where your money went
Usage Logs lists every request with its model, tokens and exact cost. Transactions shows deposits and charges. Combine that with per-key limits and you always know what you're spending.
Why prepaid?
- No surprise bills: you can't spend more than you added.
- No subscription to forget about.
- Works with payment methods that don't support recurring charges, like USDT.
Want to spend less? Read how to reduce your AI API costs.
Claude, GPT, Gemini, DeepSeek, image and video models through one OpenAI- and Anthropic-compatible endpoint. Prepaid, pay only for what you use.