iFlytek Launches Spark Medical Large Model X1, Surpassing GPT-4o and DeepSeek R1 in Clinical Reasoning

Following their breakthrough into the mainstream, large language models are being rapidly integrated across various industries, and the healthcare sector is no exception.

Not only have numerous top-tier hospitals and cutting-edge medical technology companies in China rapidly rolled out deployments, but even some local healthcare security administrations have高调 announced their adoption of this technology.

However, whether large models can truly be scaled for application in healthcare institutions ultimately depends on whether the products emerging from these technologies can address issues such as dormant data, privacy constraints, and high customization barriers, while seamlessly integrating into physicians’ workflows.

iFlytek Medical’s recently released iFlytek Spark Medical Large Model X1 may offer the best solution currently available.

As the only medical large language model with deep reasoning capabilities currently trained entirely on domestically produced computing power, the Spark Medical Large Model X1 has surpassed GPT-4o and DeepSeek R1 in performance on medical tasks such as diagnostic recommendations and health consultations. The professionalism, practicality, and personalization of its personal health assistant, iFlytek Xiaoyi, have been further enhanced.

Meanwhile, iFlytek Healthcare has partnered with Huawei to launch the “iFlytek Spark Medical All-in-One Machine” for the first time, specifically designed for medical application scenarios. By providing a secure, controllable, efficient, and intelligent integrated solution, this initiative aims to accelerate the deep empowerment of AI technologies in the healthcare sector. Furthermore, based on the iFlytek Spark Medical Large Model X1, iFlytek Healthcare has upgraded its entire product portfolio.

The Spark Medical Large Model X1 is built upon iFlytek Medical’s years of accumulation of hundreds of millions of authoritative, high-quality medical knowledge entries and hundreds of millions of high-quality, de-identified clinical diagnosis and treatment records. By integrating technological innovations in areas such as enhancement of specialized medical knowledge, evidence-based clinical reasoning, critique models for long chain-of-thought processes, tree search, and reinforcement learning from expert physician feedback, the model significantly reduces medical hallucinations. When addressing complex questions, the Spark Medical Large Model X1 can even provide step-by-step explanations of its evidence-based reasoning process.Enhancing the logical correctness, professionalism, and interpretability of reasoning in complex medical scenarios.

Deconstructing the iFlytek Spark Medical Large Model X1, its core advantages can be summarized as “efficiency enhancement” and “quality improvement,” which are also two characteristic features typically possessed by medical large models with commercialization potential.

First, improving efficiency.In medical settings, information exchange is characterized by high frequency and high density. In current clinical workflows, healthcare professionals are required to simultaneously perform multiple tasks, such as data entry and electronic health record documentation, leaving only limited time for patient observation and clinical communication.

Next is quality improvement.In the past, large medical AI models often focused on “improving efficiency” while neglecting their potential to “enhance the quality of medical services and patient experience.” However, if these models fail to expand the supply of medical services and deliver incremental value to relevant departments through quality improvement, they will likewise struggle to achieve successful commercial deployment.

The iFlytek Spark Medical Large Model X1 outperforms GPT-4o and DeepSeek R1 in reasoning tasks such as diagnostic recommendations, health consultations, and interpretation of examination and test reports, thereby further enhancing the quality of medical services and patient experience while expanding the supply of healthcare services. According to official test data, the Spark Medical Large Model X1 leads GPT-4o by approximately 10 to 20 percentage points in core metrics such as the appropriateness rate of general practice diagnoses and the accuracy of health checkup report interpretations. This advantage is even more pronounced in specialized cardiology scenarios developed in collaboration with Beijing Anzhen Hospital; for instance, its primary diagnosis appropriateness rate reaches 90.1%, significantly surpassing GPT-4o’s 75.0%.

Data Source: Real-world scenario test sets from iFlytek Xiaoyi, Zhiyi Assistant, Anzhen Hospital Department of Cardiology, etc.

In practical applications, the Spark Medical Large Model X1 has been integrated into the AI health assistant iFlytek Xiaoyi and deeply fused with personal health records. During user health consultations, it can gain deep insights into users' health conditions while anticipating potential needs, thereby providing more personalized and practically useful medical advice. This creates greater possibilities for delivering incremental value to relevant clinical departments.

In addition to enhancing model capabilities and expanding product application scenarios, iFlytek Healthcare has also innovated in deployment models for healthcare institutions and regions.

Leveraging the technological foundation of the iFlytek Spark All-in-One Machine and the Xingfei Medical Large Model X1, iFlytek Medical has launched the new “iFlytek Spark Medical All-in-One Machine,” which includes a 4U training-and-inference all-in-one unit and a 2U inference-only unit. These solutions deliver full-stack domestic support, covering computing power, models, training, inference, and applications. Designed for rapid deployment and out-of-the-box usability, they ensure that data remains within local boundaries, thereby effectively safeguarding data security.

The newly launched iFlytek Spark Medical All-in-One Machine not only continues iFlytek Medical’s deep accumulation in its technological foundation but also achieves breakthroughs in medical industry knowledge capabilities and model performance.

The Xinghuo Medical All-in-One Machine has developed 20 specialized medical AI agents, including automatic medical record generation, connotation-based quality control of medical records, intelligent follow-up, and report interpretation. Equipped with a professional large language model (LLM) toolchain, it efficiently learns from proprietary data such as private medical records and laboratory test reports, enabling healthcare institutions and regional health systems to rapidly build their own exclusive LLMs and innovative applications.

By integrating deep reasoning with industry understanding, the all-in-one machine’s medical knowledge capabilities have improved by 30%, while model hallucinations have decreased by 10%. Meanwhile, thanks to the deep integration of Huawei Ascend and iFlytek Spark on a fully localized stack, the inference performance of the iFlytek Spark Medical Large Model X1 on the all-in-one machine has also increased by 40%.

Notably, the iFlytek Spark Medical All-in-One Machine features a comprehensive toolchain—including knowledge distillation and model fine-tuning—that supports healthcare clients in building knowledge engineering frameworks and developing proprietary large language model solutions. Leveraging fully domestic computing power, the all-in-one machine improves model construction efficiency by 50%.

Furthermore, the built-in "one-stop" Spark Medical AI Agent Platform can address the needs of medical institutions and regional scenarios, supporting users based on deep thinking,Efficiently Build Healthcare-Specific AI Agents。

Meanwhile, the all-in-one device also supports the integration and optimization of the Spark Medical Large Model X1 with other models, enabling heterogeneous model collaboration and further expanding application scenarios and technological boundaries.

The “one-stop” Xinghuo Medical AI Agent Platform undoubtedly provides the healthcare industry with a novel intelligent tool, driving digital transformation and intelligent upgrading in the medical field.

Leveraging the Xinghuo Medical Large Model X1, AI has truly achieved the ability to “understand” questions, providing accurate and precise answers. It can also interact with users, engaging in multi-turn conversations like a “real person,” progressively addressing and fulfilling their diverse needs. The iFlytek Spark Medical Large Model’s adherence to a fully domestic technological roadmap has established a security barrier—from underlying architecture to practical application—for the compliant use of medical data and the protection of patient privacy.

However, these achievements are only the beginning. Leveraging the technical capabilities of the Spark Medical Large Model X1, iFlytek Healthcare has completed iterative upgrades across its full product matrix, including primary healthcare services, regional management platform solutions, and patient services. In the future, iFlytek Healthcare will continue to expand the boundaries of medical AI applications, driven by breakthroughs in foundational technologies and scenario-based innovations, aiming to create an AI diagnostic and treatment assistant for every doctor and an AI health companion for every resident.