On October 21, the National AI Application Pilot Base (Healthcare) released its first-year results in Guangzhou. Since landing in Guangdong in July 2025, the national-level base has built a matrix of 84 vertical models for healthcare, covering clinical medicine, public health and public services — medical AI is shifting from point tools to systematic delivery.
1. The data foundation behind 84 models
Guangdong leverages more than 68,000 medical institutions and 182 top-tier hospitals, accumulating 154 high-quality datasets and 381 AI healthcare application scenarios. Working with Guangzhou Laboratory, hospitals and tech companies, the base organized these resources through six trusted data spaces and provincial platforms into the 84-model matrix.
2. A hands-on deployment list, from CDC to clinics
The results are concrete: the base and Guangdong CDC completed a national shared computing platform for disease control, with 32 provinces connecting to the trusted data space and 23 completing connector deployment. The "Infectious Disease Early-Warning and Adjudication Machine 2.0" aggregates 9 types of core data and applies 55 battle-tested models; an immunization planning agent on the "Yue Miao" platform covers 65 million active users across 5,000+ vaccination sites with over 96% Q&A success; Guangdong Provincial People's Hospital has applied agents across outpatient and inpatient workflows, with its lung-cancer consultation agent deployed in 300+ hospitals nationwide.
3. Translations that cut testing costs
Two benchmark cases: a lung-cancer pathology gene multimodal model developed with the First Affiliated Hospital of Guangzhou Medical University identifies targeted genes in as fast as 1 minute, cutting costs from tens of thousands of yuan to a few hundred, with prediction accuracy up to 90%; a dermatology agent built with Guangdong Provincial Hospital of Traditional Chinese Medicine distilled 150,000 medical cases and 160,000 prescriptions, reaching over 95% diagnostic accuracy on core diseases.
Key numbers from the base's first year
· Vertical model matrix: 84 models across three directions
· Foundation: 154 datasets, 381 AI application scenarios
· Immunization agent: 65M active users, Q&A success >96%
· Gene identification: as fast as 1 min, cost cut to hundreds of yuan
The base's value lies in putting scattered hospital data, scenario demand and model capability onto one conversion pipeline. The competition in vertical medical models is moving from "point answers" to systematic delivery of data, scenarios and models — the same methodology NineZenith's Tianxing industry models apply in government and energy sectors.
(Information synthesized from Chinanews, Guangzhou Daily, Sina News and other public reports)