AI · AI Programming · AIGC

DeepSeek-v4.1-flash Joins WeChat Mini-Program’s 10-Billion-Token Free Tier

By 怕浪猫 · Nov 4, 2026

Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Developers can swap to deepseek-v4.1-flash or hy4-preview with one field edit, because the models need no console enable and tokens deduct themselves. The update also points to a broader practice: Chinese platform vendors fold model access into app development perks, not standalone API deals.

Summary

Ten billion free tokens on WeChat’s Mini-Program Growth Plan now cover three text models after the addition of hy4-preview and deepseek-v4.1-flash. The existing hy3 stays, and none of the three require manual activation in the Cloud Development console. Calls from the mini-program, cloud server, or console deduct automatically from the granted quota.

The plan suggests two main use cases: image recognition tasks get deepseek-v4.1-flash by swapping the model field and passing a cloud storage URL, while long-text jobs such as bulk document parsing or multi-turn assistant conversations use hy4-preview to skip custom context truncation. Image models Hy-image-v3.5-preview and the official Hy4 are in the pipeline, with a separate launch notice.

Any mini-program developer can apply through WeChat Public Platform’s industry capabilities menu. Each mini-program receives a free cloud environment, 10 billion tokens for text, and 100,000 image-generation credits.

Takeaways

— DeepSeek-v4.1-flash and hy4-preview join hy3, making three text models under the 10-billion-token free tier.

— No manual activation is required; token usage auto-deducts from the grant for calls across mini-program, cloud server, or console.

— For image recognition, set the model to deepseek-v4.1-flash and pass the image’s cloud storage URL.

— For long-text processing, use hy4-preview, which removes the need for context cropping and truncation logic.

— Hy-image-v3.5-preview and the formal Hy4 image model are being integrated and will be announced later.

— Developers apply through WeChat Public Platform → Industry capabilities → Mini-program Growth Plan; each mini-program gets a free cloud dev environment, 10B tokens, and 100k image generation quota.

Conclusions

DeepSeek-v4.1-flash is a third-party model, yet it gets bundled into WeChat’s free quota alongside Hy’s own models, showing a willingness to mix external and internal AI.

The ‘no extra configuration’ and auto-deduction reduce a developer’s operational load, likely to increase adoption of both models.

Hy4-preview’s positioning for long-text suggests it handles context without manual truncation, which could simplify app architecture.

Source: juejin.cn ↗ Google Translate ↗ Backup ↗