Document Processing
Quanzhen Medical needed a strong Chinese foundation model, an integrated training-and-inference toolchain and a private-delivery path so it could commercialize large-model healthcare applications for hospitals with high data-security requirements.
Reported impact
Quanzhen created a healthcare-specific large-model product line that improves medical documentation speed and quality while supporting both public-cloud and private hospital delivery.
Quanzhen Medical and Baidu AI Cloud built a two-path healthcare large-model delivery architecture around ERNIE Speed and the Qianfan platform. In the public-cloud path, Quanzhen trains and validates its domain model on Qianfan so new clinical scenarios can be tested quickly. In the private path, Baidu supplies the servers and training-inference software for a joint all-in-one appliance, while Quanzhen contributes the healthcare large model and delivers the combined system into hospitals that need stronger data control. The resulting platform supports electronic medical-record generation, pre-consultation note creation and other clinical language tasks rather than just acting as a demo chatbot. Because the model stack is wrapped with both public-cloud and appliance delivery modes, Quanzhen can serve county-level organizations and tertiary hospitals from the same product family while keeping the clinical language model, tuning loop and deployment tooling consistent.
Clinical data and language tasks are processed through the Quanzhen healthcare model on Qianfan or the private appliance, which generates EMR drafts and pre-consultation records and returns them to hospital workflows.
Baidu provides ERNIE Speed as the Chinese foundation model and Qianfan as the public-cloud training and validation platform (original: 提供Ernie Speed基座大模型……提供百度智能云千帆公有云平台).
It now generates about 150,000 medical-record drafts and 10,000 pre-consultation records per day (original: 每日生成病例草稿达到150,000份；每日生成的预问诊记录达到10,000份).
The platform generates electronic medical records in 2-3 seconds and reduces writing time by 75% (original: 2-3秒生成电子病历……75%节省病历书写时间).
The two sides built a joint all-in-one appliance for tertiary hospitals and landed it at Changzhou First People's Hospital (original: 百度智能云和全诊医学打造全诊大模型联合一体机方案……该联合一体机方案已经在常州市第一人民医院落地).
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Archived · Oct 9, 2026 Angelus Founded by a dentist whose first product was based on a patent, Angelus competes on differentiation, developing bioceramic products and fiberglass post lines that require clinical proof, intellectual property… 50,000 pages of proprietary data now powering AI-generated documents Gemini 
Cognizant Sample Modified · Aug 12, 2026 Regis Aged Care Regis Aged Care needed to compress a daily clinical handover review that could take two to three hours for 200-plus pages of notes, without missing a single resident-level clinical signal. 200+ pages of handover notes Microsoft Copilot Studio · benchmarking process · chat interface Google Cloud Sample Archived · Aug 5, 2026 DeepHealth DeepHealth needed to handle rising imaging volumes, a worsening radiologist shortage, and the need to deliver faster, more consistent cancer detection without limiting access to high-quality diagnostics to major urban… 20M+ medical images streamed yearly with Cloud Healthcare API BigQuery · Cloud Healthcare API · Gemini 
