{"slug": "two-dots-no-line", "title": "Two Dots, No Line", "summary": "Tencent Holdings released Hy3, its first full model under Chief AI Scientist Yao Shunyu, on July 6, activating 21 billion of 295 billion parameters per call and cutting hallucination rates from 12.5% to 5.4% in product evaluations. Bernstein maintained its Outperform rating and HK$780 price target, but the model does not power Tencent's WeChat-native agent Xiaowei, which reportedly uses WeLM and DeepSeek instead, leaving the commercial path through WeChat's 1.43 billion monthly active users unconnected.", "body_md": "# Two Dots, No Line\n\n### Bernstein asks if two dots make a line. Inside Tencent, they do not yet connect.\n\n** FlashPoint** is our premium quick-strike column on market-moving China tech events. Today: Tencent’s Hy3 model answers the immediate capability question but exposes an organizational one. In May, I examined\n\n[Tencent’s broad agent push](https://hellochinatech.com/p/tencent-agent-flood). In March, I traced\n\n[the tension between Tencent’s cash machine and its AI venture](https://hellochinatech.com/p/tencent-earnings-ai-spending-selloff). Now the model has arrived.\n\n**The problem is where it landed**.\n\nOn July 6, Tencent [released Hy3, the first full model release](https://www.tencent.com/tencent-hunyuan-officially-releases-hy3-advancing-agent-capabilities-and-deeper-product-integration/) under Chief AI Scientist Yao Shunyu. It took 201 days from Yao’s appointment. The model activates 21 billion of its 295 billion parameters per call, a deliberate mid-range design. In [internal WorkBuddy tests](https://www.tmtpost.com/8054278.html), task success rates climbed from 72% to 90%. Tencent reported hallucination rates in its product evaluations fell from 12.5% to 5.4%. For agent tasks, where one hallucinated step breaks the chain, reliability matters more than benchmark scores. API pricing sits at [Rmb 1 per million input tokens](https://news.qq.com/rain/a/20260706A0709C00) and Rmb 4 for output.\n\nBernstein [maintained its Outperform rating and HK$780 price target](https://wallstreetcn.com/articles/3776401) after the release. The brokerage’s framework treats Hy3 not as a frontier model but as a model sufficient to power agent commerce inside WeChat, China’s super-app with [1.43 billion monthly active users](https://static.www.tencent.com/uploads/2026/05/13/47382ae415a209fd161bc19a1f9b3704.pdf). In Bernstein’s reading, merchants will pay for AI-enhanced traffic and tools within the WeChat ecosystem, making agent-driven transaction volume the core metric. Tencent’s most valuable AI surface is WeChat. Its newest model is Hy3. Connect the two and the commercial path follows.\n\nExcept the two do not connect. What may become Tencent’s most consequential AI product, a WeChat-native agent called Xiaowei, [began limited testing in late June](https://www.thestar.com.my/tech/tech-news/2026/06/22/tencent-tests-ai-assistant-for-its-super-app-wechat-in-china). Tencent has not disclosed Xiaowei’s full model architecture. But multiple outlets, including Caixin, 36Kr, GeekPark, and Bloomberg-sourced English reporting, identify WeLM, a WeChat-team model, as the primary engine. Several also report DeepSeek handling selected tasks. None has identified Hy3 in the stack.", "url": "https://wpnews.pro/news/two-dots-no-line", "canonical_source": "https://hellochinatech.com/p/tencent-hy3-wechat-ai", "published_at": "2026-07-10 03:25:29+00:00", "updated_at": "2026-08-02 08:45:58.697807+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-agents"], "entities": ["Tencent Holdings", "Yao Shunyu", "Hy3", "Bernstein", "WeChat", "Xiaowei", "WeLM", "DeepSeek"], "alternates": {"html": "https://wpnews.pro/news/two-dots-no-line", "markdown": "https://wpnews.pro/news/two-dots-no-line.md", "text": "https://wpnews.pro/news/two-dots-no-line.txt", "jsonld": "https://wpnews.pro/news/two-dots-no-line.jsonld"}}