iFlytek Medical, a subsidiary of iFlytek, officially launched the nationally produced medical deep reasoning big model – Spark Medical Big Model X1, and demonstrated better performance than GPT-4o and DeepSeek R1 in core medical tasks such as diagnosis recommendation and health consultation. The “iFlytek Spark Medical All-in-One” released at the same time combines Huawei technology to provide medical institutions with safe and controllable intelligent solutions.

Leading performance in medical tasks, significant improvement in general diagnosis accuracy
The Spark Medical Big Model X1 is developed based on a medical knowledge base of hundreds of millions and desensitized diagnosis and treatment data, and reduces the risk of “hallucination” (generated erroneous information) through an evidence-based reasoning mechanism. In official tests, its key indicators such as the reasonable rate of general diagnosis and the accuracy of physical examination report interpretation increased by 10 to 20 percentage points compared with GPT-4o. Taking the cooperation with the Department of Cardiology of Beijing Anzhen Hospital as an example, the reasonable rate of the main diagnosis of this model reached 90.1%, significantly higher than the 75% of GPT-4o.

The model focuses on optimizing the dual requirements of efficiency and quality in medical scenarios:

Efficiency improvement: simplifying processes such as electronic medical record entry to free up medical staff time;
Quality improvement: enhancing diagnostic logic and patient service experience, for example, the AI ​​health assistant “iFlytek Xiaoyi” can provide personalized suggestions based on personal health records.

Full-stack domestic all-in-one machine landing to ensure data security and efficient deployment
iFlytek Spark Medical All-in-One Machine includes a 4U training and push all-in-one machine and a 2U inference all-in-one machine, achieving national support from computing power to application, ensuring that medical data does not leave the domain. The device has 20 built-in medical-specific intelligent bodies, covering scenarios such as medical record generation, quality control, and follow-up, supporting medical institutions to quickly build exclusive models based on private data.

In terms of technical upgrades, the all-in-one machine’s ability to understand medical industry knowledge has increased by 30%, the model hallucination rate has decreased by 10%, and the inference efficiency has increased by 40%. Through the deep integration of Huawei Ascend computing power and iFlytek technology, the model training efficiency has increased by 50%, helping medical institutions shorten the intelligent transformation cycle.

Expand application scenarios and strengthen primary medical empowerment
At present, the Spark Medical Big Model X1 has been applied to primary diagnosis and treatment, health management and other scenarios. Its multi-round dialogue and deep reasoning capabilities can simulate real doctor-patient interactions, such as gradually asking for symptom details in health consultations and providing phased guidance. iFlytek Medical has simultaneously upgraded product lines such as regional management platforms and patient service systems, and plans to continue to optimize the coverage and accuracy of AI-assisted diagnosis in the future.

iFlytek Medical emphasizes that the national domestic technology path not only ensures data compliance and privacy security, but also provides an autonomous and controllable foundation for the large-scale implementation of medical AI. With the deepening of model and hardware collaboration, the digital transformation of the medical industry will enter a more pragmatic development stage.