iFLYTEK Spark V3.7 tops MedBench 5.0

This is a notable shift from medical chatbots toward multimodal workflow agents, with deployment scale becoming as important as benchmark performance.

• –Speech, imaging, and semantic reasoning are combined for clinical documentation, diagnosis support, and patient management
• –The primary-care assistant has reportedly delivered more than 1.2 billion AI-assisted diagnostic suggestions
• –An 85% reasonable-rate for end-to-end medical-record generation points to a practical productivity gain for clinicians
• –MedBench results are company-reported, so developers should examine task definitions, safety evaluation, and real-world error rates
• –The model’s focus on China’s primary-care network gives it a deployment advantage that general-purpose model benchmarks may miss