iFLYTEK Healthcare Upgrades Spark Medical Model to V3.7, Claims 99.6 on MedBench 5.0
iFLYTEK Healthcare announced on October 22 a full upgrade of its Spark medical large model to V3.7, and reports a score of 99.6 on MedBench 5.0. The company says the model now combines speech, medical imaging and language understanding for clinical agent use. The benchmark and usage figures are company-reported.
iFLYTEK Healthcare announced on October 22 that it has completed a full upgrade of its Spark medical large model to version 3.7, and reports that the model scored 99.6 on MedBench 5.0. MedBench is the medical evaluation run by the Shanghai Artificial Intelligence Laboratory. The company says V3.7 stayed at the top of the leaderboard. The score is the company's own report.
What changed in V3.7
According to iFLYTEK Healthcare, V3.7 combines speech, medical imaging and language understanding more deeply than before. The company says this makes the model more capable of acting as an agent in clinical settings. Parameter count, context window and pricing are not yet disclosed.
Product impact
The upgrade feeds into several of the company's products, according to iFLYTEK Healthcare:
- AI clinical assistant: The rate of reasonable medical records generated by the system has risen to 85 percent. The company says this strengthens its position in primary-care general practice.
- Cloud imaging service: Report generation and quality control for hospital imaging departments have been upgraded.
- AI patient management: The system uses the stronger multimodal and evidence-based reasoning to make follow-up visits smarter and patient profiles more accurate.
- Consumer health-consultation assistant: It now keeps a dynamic memory of health records, which moves it from one-off questions toward ongoing health management.
The company now organizes its business into three areas: an AI clinical assistant, an AI health assistant and an AI digital foundation for medical institutions. It says the upgraded model is meant to support deployment at scale across all three.
Usage figures (company-reported)
- The primary-care clinical assistant covers 31 provincial-level regions and 828 districts and counties, and serves more than 77,000 grassroots medical institutions.
- It has given more than 1.2 billion AI-assisted diagnostic suggestions and helped generate more than 490 million standardized electronic medical records.
- In Anhui, the cloud imaging platform now connects every public hospital in the province and has collected more than 160 million imaging studies. The service has expanded to Guangxi and Ningxia.
- The consumer health assistant has handled more than 220 million AI consultations.
What this means
V3.7 is positioned as a model for clinical workflows rather than for question answering alone. The company ties the version to concrete product changes: record generation, imaging reports and quality control, patient follow-up, and persistent health records for consumers. The 85 percent figure for reasonable medical records is a company-reported measure for one product, and the 99.6 MedBench 5.0 score is likewise self-reported. The usage numbers show the scale at which iFLYTEK Healthcare says the Spark medical model is already deployed in primary care. Pandaily, which reported the news, notes that medical models in China are moving from single-turn question answering toward multimodal agents built into clinical workflows, especially in primary care, where doctor shortages are most acute.
Related Articles
MiniMax releases M3.1-Flash: 428B-parameter multimodal MoE with 1M-token context and weights on Hugging Face
MiniMax has published MiniMax-M3.1-Flash on Hugging Face: a native multimodal mixture-of-experts model with about 428B total parameters, 23B activated per token, and a context window of up to 1M tokens. It builds on MiniMax-M3 with sparse attention, quantization-aware training and DSpark speculative decoding.
iFLYTEK releases Spark X2.6, a 293B-A30B MoE model trained on domestic compute
At its 1024 developer festival, iFLYTEK released Spark X2.6, a 293B-A30B mixture-of-experts model trained on domestic compute. The company also open-sourced the on-device Spark X2.6-4B and debuted the SparkAudio 1.0 speech model.
inclusionAI releases Ling-3.0-flash-Sante, a 124B healthcare model with 5.1B active parameters
inclusionAI has released Ling-3.0-flash-Sante, a healthcare-specialized model built on Ling-3.0-flash with 124B total and 5.1B active parameters per token. Weights are on Hugging Face and ModelScope, and training extends to 256K tokens. The company labels it a research model, not a medical device.
Kyutai releases CADRE, 7 open causal diffusion video autoencoders for interactive world models
Kyutai released CADRE, a family of 7 open causal diffusion video autoencoders built for interactive world models. World models built on CADRE run at 20 fps on 1 H100 and react in under 230 ms, according to MarkTechPost's report.
Comments
Loading...