The AI resource pack included in the Mini Program Growth Plan now covers two additional models: hy4-preview and deepseek-v4.1-flash. Together with the existing hy3, the 1 billion text-generation tokens now support three text models.
None of the three models need to be enabled in the CloudBase console. Whether you call them from the Mini Program client, the CloudBase server side, or the CloudBase console, token usage is automatically deducted from the granted quota, with no extra configuration.
How to use them in your Mini Program
Use the table below to map the new capabilities to your use case:
| Recommended scenario | What to change |
|---|---|
| Image recognition (receipt reimbursement, automatic grading of scanned exam papers, product screenshot classification, extracting tables from images) | Set model to "deepseek-v4.1-flash" and pass the Cloud Storage URL of the image directly. |
| Long-context processing (batch parsing of long documents, summarizing long support conversations, multi-turn assistant dialogues) | Set model to "hy4-preview" and drop the tedious context trimming and truncation logic. |
The image generation model Hy-image-v3.5-preview and the newly released Hy4 are also being integrated; we will announce their availability separately.
How to claim the quota
All Mini Program developers can apply. A single Mini Program can claim a free CloudBase environment, 1 billion text-generation tokens, and 100,000 image-generation tokens under the program rules. Hy4 Preview, DeepSeek and other designated models are now supported — try them in the Mini Program Growth Plan.
Where to apply: WeChat Official Accounts Platform → Industry Capabilities → Mini Program Growth Plan.
What kind of Mini Program are you building? Or what problems have you run into? Join the conversation in the comments.
