Launch of Lianhui Technology's Second-Generation OmAgent: A Cutting-Edge Multimodal Intelligent Agent

During the World Artificial Intelligence Conference (WAIC), Lianhui Technology unveiled the second generation of its multimodal intelligent agent, OmAgent. Compared to the first generation released last year, this upgraded model boasts significant improvements in its perception module and decision-making capabilities.

The perception module has received a comprehensive enhancement with the introduction of OmDetV2, which accelerates the overall sensing of the environment. This advancement redefines the fundamental architecture of perception, launching the EFH high-performance fusion head. This innovative setup incorporates various optimization techniques for model acceleration, language vector caching, and lightweight feature encoding and decoding.

In terms of decision-making, Lianhui Technology has introduced the new OmChatV2, a second-generation large language model based on native multimodal pre-training. This model is available in configurations of 8B, 40B, and 60B parameters, accommodating diverse requirements. It adeptly supports a variety of complex inputs, including video, images, and text, making it ideal for the intricate scenarios that intelligent agents often encounter.

To ensure practical applicability, Lianhui Technology has also completed compatibility and performance validations with various Chinese GPUs. The company has made remarkable strides in multimodal intelligent agent technology. The OmAgent framework integrates comprehensive modules such as perception, memory, and decision-making, while incorporating capabilities from different large models like OmDet and OmChat. This integration simplifies application development for businesses and developers, propelling intelligent agent technology into deeper and broader domains.

With OmAgent, complex problems across various scenarios can be resolved swiftly and accurately. To facilitate the integration of intelligent agents into daily work and life, Lianhui Technology has launched a new series of multimodal intelligent agent products, including the Spatial Operation Agent and Knowledge Service Agent, designed to act as "super assistants" for industry users.