Per-request API quickstart
Send a request and pay for a successful result. Choose a dedicated workspace when you need your own tools, custom nodes or a different workflow.
Choose the charging model
Per request: use a model listed below for a supported task. You do not manage a dedicated GPU.
Workspace: choose your model, hardware and current hourly quote. Billing includes startup and continues until you stop it. Explore workspaces →
Compare your expected request volume with the hours a dedicated machine would run; there is no universal cheaper option.
Start with one request
- Create an API key and add prepaid credit in Billing.
- Choose a configured model and review its price below.
- If the response contains
poll_url, poll that URL for the saved job status. A successful HTTP submission is not a completed result.
curl -X POST https://modelpilot.ai/api/v1/generate/image \
-H "Authorization: Bearer mp_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{"model":"flux-schnell","prompt":"a red apple on a white table"}'Configured models and prices
Rates come from the same configuration used for successful-request billing. A model listing does not reserve GPU capacity. Cold startup figures are configured estimates, not delivery promises; processing adds time and depends on your input.
Image generation
| Model | Request model | Successful request | Cold startup estimate |
|---|---|---|---|
| Flux Schnell | flux-schnell | $0.0170 | ~30s |
| Flux Dev | flux-dev | $0.0170 | ~30s |
| Stable Diffusion XL | sdxl | $0.0140 | ~25s |
| Z-Image Turbo | zimage | $0.0140 | ~25s |
Speech & audio
| Model | Request model | Successful request | Cold startup estimate |
|---|---|---|---|
| Kokoro TTS | kokoro | $0.0090 | ~15s |
| Chatterbox Turbo | chatterbox | $0.0170 | ~30s |
Text & reasoning
| Model | Request model | Successful request | Cold startup estimate |
|---|---|---|---|
| Qwen3 4B | qwen3-4b | $0.0140 | ~150s |
| Qwen3 8B | qwen3-8b | $0.0190 | ~240s |
| DeepSeek R1 8B | deepseek-r1-8b | $0.0190 | ~240s |
| GLM-Z1 9B | glm-z1-9b | $0.0190 | ~270s |
| DeepSeek R1 14B | deepseek-r1-14b | $0.0470 | ~420s |
| Qwen3 32B | qwen3-32b | $0.1380 | ~720s |
| GLM-Z1 32B | glm-z1-32b | $0.1380 | ~720s |
For guided video creation, replacement or extension, check Studio for the methods currently offered and their limits. This table does not imply a video API is available.
What changes the wait?
- A cold worker: model download and loading can dominate the first request.
- A warm worker: startup can be avoided, but generation time still depends on the model and input.
- Idle scale-down: the application configures an idle timeout capped at 60 seconds. A later request can need another startup.
- Capacity: configured GPU candidates differ by model. A published model and a price do not guarantee a free worker.
Budget example
At the configured Flux Schnell rate, 10 successful requests cost $0.1700. At that same rate and volume for 30 days, the requests total $5.1000. This is request-price arithmetic, not a forecast of traffic or latency.