Tencent just gave away one of the biggest AI models anyone has built, for free.
The company open-sourced Hy4 preview, a model built on 770 billion parameters. It only lights up 49 billion of those for any given task, which is the trick that keeps it fast and affordable instead of impossibly expensive to run. The headline number, though, is the context window: over a million tokens, and it gets there natively instead of stretching a smaller model past its limits with tricks. That means you can hand it enormous amounts of text, code, or data in one go and it actually holds onto all of it.
Tencent didn’t just publish a spec sheet and call it a day. They ran a blind test with 163 outside engineers grading 203 real coding and engineering tasks. Hy4 preview came out ahead of two of the strongest open rivals around, GLM 5.3 and Kimi K3, on average scores. The margin was small, but a win graded by people who didn’t know whose model they were scoring still counts.
You can already get your hands on it. It is live inside Tencent’s WorkBuddy and CodeBuddy apps, both the Chinese and international versions, plus Yuanbao and ima, and the API is up on Tencent Cloud and OpenRouter for well under a dollar per million input tokens.
Here is the part that matters for the rest of us. Every time a lab open-sources something this capable, the price of a genuinely good model drops for everyone downstream, including the small teams and solo builders who could never afford to train one themselves. The race between Chinese and US labs gets framed as a scary thing a lot. Days like this are a reminder it also means better tools land in more hands, faster than they used to.
——
Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
——
Sources:
https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/
https://technode.com/2026/08/28/tencent-open-sources-hy4-preview-with-770b-parameters-and-a-1m-token-context/
https://recipes.vllm.ai/tencent/Hy4-preview
Repost this. Thanks.
