Model Calls
Official Models
Official models are services published by the platform or its partners, providing a more stable and efficient experience. The pricing of model calls varies depending on the task type. Text dialogue models are billed based on token input/output, while image and video generation models are billed per generated image or video.
Text Dialogue
| Model Name | Billing Unit | Input Price (RMB) | Output Price (RMB) |
|---|
| GpuGeek/DeepSeek-R1-671B | /Million Tokens | 8 | 32 |
| GpuGeek/DeepSeek-R1-Distill-Llama-70B | /Million Tokens | 2 | 8 |
| GpuGeek/DeepSeek-R1-Distill-Qwen-32B | /Million Tokens | 1.5 | 6 |
| GpuGeek/DeepSeek-R1-Distill-Qwen-14B | /Million Tokens | 0.6 | 2.4 |
| GpuGeek/DeepSeek-R1-Distill-Llama-8B | /Million Tokens | 0.4 | 0.8 |
| GpuGeek/DeepSeek-R1-Distill-Qwen-7B | /Million Tokens | 0.4 | 0.8 |
| GpuGeek/DeepSeek-R1-Distill-Qwen-1.5B | /Million Tokens | 0.2 | 0.4 |
| GpuGeek/qwen2.5-0.5B | /Million Tokens | 0.1 | 0.2 |
Text-to-Image
| Model Name | Billing Unit | Output Price (RMB) |
|---|
| GpuGeek/Stable-Diffusion-3.5-Large-Turbo | /Image | 0.15 |
| GpuGeek/Cogview4-6B | /Image | 0.15 |
Text-to-Video
| Model Name | Duration & Resolution | Billing Unit | Output Price (RMB) |
|---|
| Vidu/text2video-1.5 | 4S360P | /Video | 1.25 |
| Vidu/text2video-1.5-HD | 4S720P | /Video | 3.13 |
| Vidu/text2video-1.5-FHD | 4S1080P, 8S720P | /Video | 6.25 |
Image-to-Video
| Model Name | Duration & Resolution | Billing Unit | Output Price (RMB) |
|---|
| Vidu/image2video-2.0 | 4S720p | /Video | 1.25 |
| Vidu/image2video-2.0-HD | 4S1080p, 8S720p | /Video | 3.125 |
Personal Public Models
Personal published public models are billed based on runtime duration per call. The cost per unit time depends on the hardware chosen by the publisher. Currently supported hardware type:
| Hardware Type | Billing Unit | Price (RMB) |
|---|
| RTX-4090-24G | /Card | 2.18/hour |
Private Models
Private models are billed based on the online duration of the hardware configured for the model. Online billing includes startup time, request processing time, and idle time after startup. When the model receives no requests, it will switch to a "cold start" state in about 10 minutes, no longer occupying hardware or incurring costs.
| Hardware Type | Billing Unit | Price (RMB) |
|---|
| RTX-4090-24G | /Card | 2.18/hour |
Model Deployment
| Hardware Type | Billing Unit | Price (RMB) |
|---|
| RTX-4090-24G | /Card | 2.18/hour |