Tencent open-sources Hunyuan Hy4 preview — 770B params, 1M context

Share
Tencent open-sources Hunyuan Hy4 preview — 770B params, 1M context

China's open-weight race just got its heaviest new contender, and Tencent is pitching it as a productivity workhorse rather than a benchmark-chaser.

Tencent is open-sourcing Hy4 preview, a 770 billion-parameter Mixture-of-Experts model with a 1 million-token context window.

The architecture activates only 49 billion parameters per token — a 78-layer stack where each layer beyond the first holds 256 routed experts plus one shared expert, with eight experts firing per token. It ships under the Apache 2.0 license with an FP8 quantized version included, and is live on Hugging Face, ModelScope, and GitCode, with vLLM and SGLang support plus access through Tencent Cloud TokenHub and OpenRouter. It is the follow-up to April's Hy3 preview (295 billion total parameters), and Tencent telegraphed it on its August 14 earnings call.

Under the hood it leans hard on inference efficiency: a gated take on DeepSeek's sparse attention (Gated DSA) with IndexCache reusing sparse indices across layers, identity-hyperconnection residual paths, and a native multi-token-prediction layer for speculative decoding. The stated targets are software engineering's long-horizon tasks, cross-file office work, one-prompt playable game prototypes, and research reasoning.

The benchmark story is the most interesting part, and it requires careful reading. Tencent published no public benchmark scores. Instead it ran 163 internal experts across 203 engineering tasks, scoring 2.99 out of 4.00 — marginally ahead of GLM-5.3 (2.92) and Kimi K3 (2.94). That is a net win of roughly six percentage points over GLM-5.3, real but thin enough that the honest verdict is "same tier, third-party confirmation pending" — especially since Tencent set the questions and did the grading itself.

Tencent is transparent about the caveats: it calls this an early preview, acknowledges the model still underperforms on long reasoning and tends to over-verify itself, and its science demos — a 2× speedup on a molecular-dynamics simulation, quantum-transport leakage cut from 48.2% to 4.8%, and a century-old geometry lower bound pushed from 0.3808 to 0.41104 — are un-peer-reviewed and were run in Tencent's own environment.

What to watch: independent benchmarks are the only thing that will confirm whether Hy4's internal lead survives contact with evaluators Tencent doesn't control.

Does a 770B model shipping without public benchmark scores deserve first-tier status yet? Write to us via our contact page.

Sources: Tencent-Hunyuan/Hy4-preview (GitHub)

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Qwen3.8-27B hits 1M downloads as small open models close the frontier gap

Qwen3.8-27B hits 1M downloads as small open models close the frontier gap

Alibaba's latest open-source release is turning heads: Qwen3.8-27B, a 27-billion-parameter model that runs on personal computers, has crossed one million downloads within days of launch and is outperforming models several times its size on independent benchmarks. The model scored 52 on the Artificial Analysis Intelligence Index — placing it above every medium-class open model (40B–150B parameters) and tying with DeepSeek V4 Flash, which ranks fifth among all large models.

Zhipu's GLM-5.3 ships with 'emergent' cyber capabilities

Zhipu's GLM-5.3 ships with 'emergent' cyber capabilities

China's open-weights race just produced its most awkward win: Zhipu's GLM-5.3 landed Thursday with frontier-level coding and a cyber capability its own makers admit outgrew their expectations — and Moonshot AI is racing a closing window toward a Hong Kong listing. Zhipu released GLM-5.3 on Thursday, an open-weights flagship whose every gain comes from post-training — including an "emergent" cyber capability the company says developed faster than it expected.

Qwen open-sources Qwen3.8-2.4T-A95B, its first Max-class model

Qwen open-sources Qwen3.8-2.4T-A95B, its first Max-class model

Wednesday's wire: Alibaba opened the weights on its most powerful model ever, Honor put a robotic arm in a phone and sold 400,000 of them before lunch, and Europe's booksellers started refusing suspicious bulk orders. Alibaba's Qwen team open-released Qwen3.8-2.4T-A95B on Wednesday — the first Qwen-Max-class model to ship as open weights, with 2.4 trillion total parameters and 95 billion active in a 512-expert mixture-of-experts architecture.

Mistral unveils Le Chonk: a 1T-parameter open-weights model

Mistral unveils Le Chonk: a 1T-parameter open-weights model

The biggest open-weight release outside China lands in public preview today, and the country that spent the week promising its own frontier model just put a price on the ambition. Mistral has opened a public preview of Mistral Large 4 — codenamed "le Chonk" — a 1 trillion-parameter mixture-of-experts model with 49 billion active parameters, natively multimodal, which the company calls its largest and most capable model to date.

Arete-27B: Nous cut Qwen's propaganda rate from 22% to 0.9%

Arete-27B: Nous cut Qwen's propaganda rate from 22% to 0.9%

Nous Research took Alibaba's Qwen3.8-27B, put about fifty million tokens of supervision through it, and shipped back a model that answers political questions like a Western lab's build. The interesting part is how little it took — and that the whole thing is checkable. Nous Research has released Arete-27B, an Apache 2.0 model post-trained from Alibaba's Qwen3.8-27B, and says the share of answers containing Chinese state propaganda fell from 22 percent to 0.9 percent while general capability…