Tencent Yuanbao AI Assistant has recently introduced a long-form intensive reading feature, marking another new advancement following the multi-document and ultra-long-form extensive reading summaries. Users who upload professional content such as papers, financial reports, or research reports can not only receive text summaries but also enter a deep reading mode. This mode offers an overview of core content, modular analysis, and summarizing charts to help users quickly grasp key information.
The deep reading mode leverages Tencent's Hunyuan large model's capability to handle long texts, natively supporting inputs of up to nearly 500,000 words, and generating content rich with illustrations, including original image citations and advanced analytical charts created with code. It can also extract innovation points and shortcomings from papers, aiding users in evaluating paper quality, and provides a search or translation function for technical terms.
For financial reports and research reports, Tencent Yuanbao can organize them from multiple dimensions, generating professional charts such as DuPont analysis, and ensuring the accuracy of numerical calculations through an integrated calculator function. Users can also view uploaded and deeply read files in the "My Files" sidebar on the homepage, and even review the intensive reading content and original text in offline mode.
Tencent Yuanbao is based on Tencent's Hunyuan large model, which has expanded its parameters to a trillion-scale, adopting a Mixture of Experts (MoE) structure with multi-modal understanding and generation capabilities. Tencent Hunyuan ranks first in the SuperCLUE-V benchmark for multi-modal understanding among Chinese large models, demonstrating outstanding multi-modal understanding abilities. Additionally, Tencent Hunyuan's text-to-image large model has been fully open-sourced for free commercial use by enterprises and individual developers.