b1banner01.jpg ph-173935156054.jpg

Yan Architecture Model

China's first LLM with non-transformer architecture or non-attention mechanism

  • Novel Architecture

  • Comprehensive Adaptation

  • Collective Intelligence

Collective Intelligence

b1_img05.pngb1_img04.pngb1_img03.pngb1_img02.pngb1_img01.pngb1_img06.pngb1_img07.pngb1_img08.png

Product Version

  • Yan-SDK Std

    Widely Applicable Standard Model

    • b1_icon01.svg

      Multimodal capabilities for text, image, and speech

    • b1_icon02.svg

      Stability and reliability

    • b1_icon03.svg

      Broad device compatibility

    Hardware compatibility: Mid-to-low performance CPUs

    Designed to meet the cost-efficiency and smooth performance requirements of mid-to-low-end smartphones, PCs, and wearable devices.

  • Yan-SDK Pro

    High Performance Professional Model

    • b1_icon01.svg

      Multimodal capabilities for text, image, speech, video

    • b1_icon05.svg

      Real-time outstanding performance

    • b1_icon06.svg

      Comprehensive capability interface

    Hardware compatibility: High-performance CPUs、NPU

    Designed to meet the high-performance and customization needs of UAV, humanoid robots, and professional devices.

Model Capabilities

  • b1_icon07.svg

    Machine Understanding

    Relying on powerful comprehension capabilities, the system can deeply analyze and integrate information from multiple modalities. This ability enables the system to understand complex scenarios and tasks more comprehensively, thereby making more accurate and intelligent decisions.

    01/04

    b1_img06.jpg
  • b1_icon08.svg

    Modality Recognition

    Accurately identifies user intent expressed through multiple modalities such as text, image, and speech. This capability enables the system to understand user needs more precisely, thereby providing more personalized services and responses.

    02/04

    b1_img07.jpg
  • b1_icon07.svg

    Intellingent QA

    Through the integration of extensive knowledge and voluminous data, the system is empowered to carry out complex knowledge reasoning and question-answering (QA) tasks. This remarkable capacity enables it to adeptly respond to diverse types of questions, spanning professional knowledge inquiries and common-sense questions, providing users with all-encompassing information support.

    03/04

    b1_img02.jpg
  • b1_icon08.svg

    Instruction Execution

    Upon accurately discerning users' intents, the system is capable of promptly executing the corresponding instructions. Whether through text-based commands, visual image-based instructions, or voice-activated controls, the system can swiftly respond and complete tasks,thus improving work efficiency and user experience.

    04/04

    b1_img04.jpg

Model Performance

英(1).png

Model Advantages

  • b1_icon11.svg

    Novel Architecture

    Non-transformer architecture Brain-like activation mechanism

  • b1_icon12.svg

    Comprehensive Adaptation

    Widely compatible with Raspberry Pi, robots, uavs, smartphones, PCs, and other end-side devices

  • b1_icon13.svg

    Collective Intelligence

    Self-directed Learning,continuous evolution,Strong interaction, high collaboration

  • b1_img11.jpg
  • img12.jpg
  • b1_img05(1).jpg
syforwa.jpg ph.jpg

Win-Win Cooperation

Make every device its own intelligence

a0icon15.svg

Book Now