Hy4 Preview is a 770B MoE model (49B active, 1M context) from Tencent, built for long-horizon agentic tasks. It autonomously handles coding, game dev, and complex document analysis, running its own tests and fixing bugs before delivery.
Hy4 preview is the next step after Hy3, and Tencent went much bigger this time. 770B total, 49B active, 1M context.
Tencent built a lot of the training around real work from its own engineering and specialist teams. The model is meant to stay with a task when it gets long or messy.
They also ran a blind eval with 163 internal experts across 203 engineering tasks. Hy4 preview came out slightly ahead of GLM 5.3 and Kimi K3 🤔
the overthinking bit is what i'd watch @zaczuo , more than the eval numbers. on long jobs the thing that kills you isn't being wrong, it's burning the window re-checking stuff it already settled, and 1M of room makes that easier to do and harder to spot. 49B active is what decides if anyone runs it outside a datacenter, but 770B total is still 770B you havee to park somewhere.
About Hy4 preview on Product Hunt
“Tencent’s 770B open model for long-horizon work”
Hy4 preview launched on Product Hunt on August 29th, 2026 and earned 180 upvotes and 3 comments, earning #2 Product of the Day. Hy4 Preview is a 770B MoE model (49B active, 1M context) from Tencent, built for long-horizon agentic tasks. It autonomously handles coding, game dev, and complex document analysis, running its own tests and fixing bugs before delivery.
Hy4 preview was featured in Open Source (68.8k followers) and Artificial Intelligence (477.2k followers) on Product Hunt. Together, these topics include over 132k products, making this a competitive space to launch in.
Who hunted Hy4 preview?
Hy4 preview was hunted by Zac Zuo. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Hy4 preview stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Zac Zuo
Hi everyone!
Hy4 preview is the next step after Hy3, and Tencent went much bigger this time. 770B total, 49B active, 1M context.
Tencent built a lot of the training around real work from its own engineering and specialist teams. The model is meant to stay with a task when it gets long or messy.
They also ran a blind eval with 163 internal experts across 203 engineering tasks. Hy4 preview came out slightly ahead of GLM 5.3 and Kimi K3 🤔
It is open under Apache 2.0, and you can try it now through OpenRouter or @WorkBuddy.
They are still calling it a preview. It overthinks and over-checks itself. Same habit as Hy3: put it out, see what breaks!