Tencent opens Yuanbao Search API: 100B-scale index, ~0.42s average latency
智东西 · wechat · 2026-10-21
Tencent has upgraded Yuanbao Search and opened it to developers via Tencent Cloud APIs, positioning it as the "water and electricity" of the AI industry. The service now indexes hundreds of billions of pages across 13 verticals with 157 authoritative sources, serving hundreds of millions of daily calls across 50+ internal Tencent products.
Key facts from hands-on testing and disclosure:
- Three integration paths (Skill, MCP, API); free tier of 2,000 calls/month, up to 6,000/month with invite codes through November.
- API latency measured at 345-485ms per call (0.42s average); results include structured cards for weather, stock prices and gold prices, letting agents skip crawling and parsing.
- Custom disk vector index fused with inverted index lifted recall from under 80% to over 93% while cutting machines from 3,200 to 400; GistToken compression optimizes inference cost.
- In WorkBuddy, average search calls per task dropped from 8 to 5, satisfaction rose 10-20%, and fastest responses are now under 300ms.
Unlike OpenAI and Google embedding search inside model APIs, Tencent offers search as a standalone service, giving developers model choice and workflow control. Source governance filters AI-generated low-quality content and blacklists repeat offenders.
More from coding & agent
- Testing Hermes Agent with n8n: Letting Agents Execute Real Automation Workflows — techlatest_net · 2026-10-23
- OpenAI: Keep MCP Server Instructions Under ~1,000 Characters or ChatGPT May Truncate Them — bibryam · 2026-10-23
- Only ~56% of MCP Servers Provide Instructions, OpenAI Data Shows — bibryam · 2026-10-23
- OpenAI: Don't Hide MCP Tools Behind listCommand/executeCommand Dispatchers — bibryam · 2026-10-23
- Resend logs 1M monthly MCP calls as OpenAI frames MCP as a distribution channel — bibryam · 2026-10-23
- MCP Elicitation lands in Claude: servers can ask users questions mid-action — bibryam · 2026-10-23