Tencent Open-Sources Hy4 AI Model to Transform Complex Financial Workflows

Tencent has launched Hy4, its latest large language model, positioning it as a top-tier open-source tool for the global AI market. For fintech professionals, the model’s specialized training in financial analysis and its new "computer-use" capabilities represent a significant shift toward autonomous, agentic workflows in fund research, quantitative investment, and securities analysis.

What was announced

Hy4 is a next-generation large language model that introduces visual understanding and computer-use capabilities. Unlike previous iterations, Hy4 can interact with command-line and graphical user interfaces, allowing it to perform tasks across professional software like Blender and CAD. Tencent plans to release the model weights and system cards shortly, focusing on cybersecurity and biological safety as part of its open-source commitment.

The model has been specifically optimized for the financial sector. It was trained on thousands of specialized workspaces and tens of thousands of tasks derived from real-world scenarios, including fund research and quantitative investment strategies. In a blind evaluation conducted with Tencent Financial Technology, 213 financial professionals—including 153 experts from external institutions—rated Hy4 as the top-performing open-source model across 579 real-world tasks. Approximately 70% of these evaluators had over five years of industry experience.

Technical improvements include a reduction in the hallucination rate from 12.95% in the Hy4 Preview to 7.2%. Errors stemming from unsupported assumptions also saw a sharp decline, falling from 23.1% to 5.6%. These refinements are supported by targeted search optimization designed to verify factual consistency and integrate multiple information sources in real-time. The model also features enhanced instruction following and long-context comprehension, supporting fine-grained image processing and mathematical visual reasoning.

The companies involved

Tencent is a global technology conglomerate headquartered in China, operating a vast ecosystem that spans social media, gaming, and financial services. The company is perhaps best known in the fintech space for its Weixin Pay platform, which dominates the Chinese mobile payments landscape alongside competitors like Alipay. Beyond consumer payments, Tencent has established a significant presence in cloud computing and enterprise software through Tencent Cloud, providing infrastructure and AI tools to businesses worldwide.

The firm’s financial technology arm, Tencent Financial Technology, plays a critical role in developing sophisticated tools for the banking and investment sectors. This division’s involvement in the testing and refinement of the Hy4 model underscores Tencent’s strategy to move toward vertically integrated solutions for high-stakes industries. With a history of 20 stories covered by FF News, the company remains a central figure in the intersection of big tech and financial services, leveraging its massive data sets to train models that understand complex market dynamics and professional workflows.

What FF News has reported before

FF News has tracked Tencent’s expanding influence in both the payments and cloud infrastructure sectors. Recently, the company has focused on the evolution of autonomous financial tools, as seen when Tencent signs Mastercard MoU on agentic payments, days after a similar Visa deal. This move into agentic payments aligns with the computer-use capabilities introduced in the Hy4 model.

Additionally, the company’s infrastructure reach was highlighted when Simpaisa Collaborates with Tencent Cloud to Accelerate Fintech Cloud Transformation. On the consumer side, the firm continues to bridge international payment gaps, recently noted when PayPal World and Weixin Pay Enable Seamless QR Payments for US Travelers in China. These developments illustrate a broader trend of Tencent integrating its AI and payment ecosystems to serve a global audience.

What this means

The release of Hy4 signals that the "agentic" era of fintech is moving from theoretical research to practical application. By enabling an AI to interact directly with professional software and graphical user interfaces, Tencent is challenging the traditional boundaries of financial automation. The pressure is now on Western AI labs and legacy financial software providers to match this level of specialized, open-source performance.

The significant reduction in hallucination rates addresses the primary barrier to AI adoption in regulated industries: reliability. However, as these models gain the ability to execute tasks autonomously, the industry must grapple with the oversight of AI agents that can now "act" rather than just "advise." This development raises critical questions about how much autonomy institutions are willing to grant open-source models in quantitative environments.