Hy4 keeps its list prices, while TokenHub’s half-price offer has an end date
Tencent launched Hy4 on November 4. List prices match Preview, and Tencent Cloud TokenHub cuts them in half through December 31, which puts input at about US$0.90 per million tokens and output at US$2.70, with cached input near US$0.045.
Hy4 runs now on Tencent Cloud TokenHub, WorkBuddy, CodeBuddy, Yuanbao, WeCom’s AI Assistant and ima. Third-party platforms such as OpenRouter serve it too. It adds visual understanding and computer use, both absent from Preview. WorkBuddy and CodeBuddy are free until November 15.
OpenRouter’s Hy4 endpoint is Tencent’s own. Context is 1,048,576 tokens, output caps at 65,536, and the listing shows US$0.417 per million input and US$1.2505 per million output.
Tencent ran the financial evaluation itself. It asked 213 financial professionals to judge 579 real-world tasks, and Hy4 scored 3.04 against 2.97 for DeepSeek-V4.1-Flash-1017. The weights are not out. Tencent says they and the system cards follow in the coming days, with no date.
A price set from Hy4’s discount may not survive December 31
A product price set from today’s bill leaves the margin after December 31 untested. Write the most expensive call path twice instead: discounted rate, then list price. The cost you subtract from revenue belongs to the endpoint you call. TokenHub’s offer ends December 31, and that date says nothing about OpenRouter’s prices. Test the product against Tencent’s published figures, US$0.90 and US$2.70.
The rest of today’s news
- Ripple-65B Preview released: Apache 2.0 weights are out, but OpenRouter’s hosted context is 16,384 tokens, not the 1M the model card describes. Hugging Face
- OpenAI to acquire Pierre: Code Storage features move into Codex after closing, and existing customers can keep using the service. OpenAI
- Gemini 3.8 TTS voice remixing: The Gemini API release notes add natural-language instructions that turn an existing voice into an independent stored voice you can reuse by ID. Gemini API release notes