Tencent launches Yuanbao Search, a standalone AI search product at so.qq.com
Tencent Cloud announced Yuanbao Search on November 11, a standalone AI search product reachable at so.qq.com and selectable as a search engine in the latest QQ Browser. Tencent says it rebuilds the full search pipeline on the large-model capabilities behind its Yuanbao assistant and draws on an index of hundreds of billions of pages.
Tencent has launched Yuanbao Search, a standalone AI search product, Tencent Cloud announced on November 11. It is available by typing so.qq.com into a browser's address bar, and users of the latest QQ Browser can select it as their search engine in settings.
"Yuanbao Search" is a direct translation of the Chinese name. The product page is in Chinese only, and Tencent's official English name for it could not be found. The announcement was reported by IT Home, a Chinese-language outlet, and by Pandaily in English.
What Tencent says it built
According to Tencent, Yuanbao Search rebuilds the whole search pipeline on the large-model capabilities behind its Yuanbao assistant. That covers query understanding, retrieval, ranking and how answers are assembled.
- Index: hundreds of billions of pages.
- Coverage: the open web, Tencent ecosystem content (WeChat Official Accounts, WeChat Channels and Tencent News), and academic literature.
- Source priority: Tencent says results favor authoritative institutions and official sources. It reports 157 such sources integrated, with citations shown.
- Ads: Tencent says ads are clearly separated from answers. A "collapse ads" option folds the whole ad block away.
- Academic search: the product summarizes a paper's conclusions in plain language while keeping the original evidence. It also translates English titles and abstracts with one click.
- Input: users can search by typing, speaking or taking a photo, and can ask follow-up questions in an AI mode.
Tencent also says the product uses a recursive self-improvement mechanism that finds problems from real usage and feedback and improves over time. Tencent did not describe how the mechanism works or what safeguards apply.
Not disclosed
Tencent did not say which underlying models power the answers. It did not say whether ads can appear inside AI-generated answers, or whether a standalone mobile app or availability outside China is planned. Usage figures were not disclosed. Pricing is not applicable to the announcement as described, and no benchmark results were published.
The claims about source priority and ad separation come from Tencent and its product page. They have not been independently tested.
What this means
Yuanbao Search gives Tencent's search capability its own entry point, separate from the Yuanbao assistant, while Tencent says it runs on the same large-model capabilities. The two access routes, a direct URL and a search-engine option in QQ Browser, are the only distribution channels Tencent has described.
The design choices Tencent highlights are citations, a stated list of 157 official sources, and a collapsible ad block. These are product claims that users can check for themselves. Whether source prioritization holds up on contested queries, and how ads behave inside AI mode, are open questions because Tencent has not addressed ads in generated answers.
The inclusion of WeChat Official Accounts, WeChat Channels and Tencent News in the index is a notable part of Tencent's description. Content from those sources is part of what the product says it searches, alongside the open web and academic literature.
For developers and researchers, the missing model details and the Chinese-only product page limit what can be evaluated today. Neither an API nor an English-language product description has been announced.
Related Articles
Meta's Muse AI agent gets its own @muse.ai email address, paid users first
Meta's personal AI agent Muse can now be assigned a dedicated email address ending in @muse.ai, starting November 11 (Japan time). It is available first to paid users, with a rollout to all users, including free ones, planned later.
Anthropic makes Claude Enterprise inference hooks generally available, adds Security Center and EDR integration in beta
Anthropic announced on November 5, 2026 an expansion of Claude Enterprise security and admin features. Inference hooks for DLP are generally available, while Security Center, a parent-organization console, Admin API sign-out and EDR integration are in beta.
Zed v1.29.1-pre adds Claude Fable 5.2 and Claude Mythos 5.2 support, plus Git Panel branch mode
Zed Industries tagged v1.29.1-pre, which adds Claude Fable 5.2 and Claude Mythos 5.2 to the Anthropic provider. It also adds a Git Panel branch mode, a bookmark picker, configurable Git date formats and a setting to disable automatic agent thread titles.
Google launches Gemini Decisions API, handling up to 1,000 independent questions per request
Google introduced the Gemini Decisions API on November 11, 2026. The endpoint, powered by Gemini Decision 1, evaluates as many as 1,000 independent questions across text, images, audio, video and PDFs in a single request.
Comments
Loading...