Published on

YuanShi Intelligence Secures Funding for AI Model Development

Authors
  • avatar
    Name
    Ajax
    Twitter

YuanShi Intelligence's Funding and Company Overview

YuanShi Intelligence, a standout alumnus of the Chiji Startup Camp in Spring 2024, has officially announced the completion of its multi-million RMB angel round financing on December 25, 2024. Founded in June 2023, the Shenzhen-based company focuses on cutting-edge research in large model architectures and AI applications. As a rising star in the AI field, YuanShi Intelligence has garnered continuous recognition from the capital market due to its innovative technology and clear development strategy. Prior to this angel round, the company secured a seed round led by Chiji Venture Capital in January 2024, demonstrating strong market confidence in its technical capabilities, growth prospects, and team execution. This also indicates the company's significant development potential and market space in the AI sector.

Funding Allocation and Strategic Plans

The newly acquired funds will be primarily directed towards three core areas:

  • Accelerating the Evolution of the RWKV Architecture:
    • Increased investment in the underlying technology of the RWKV architecture to continuously improve model performance, efficiency, and stability.
    • Expansion of the R&D team to explore multi-modal fusion, improve and optimize RWKV multi-modal models, and broaden their application scope.
    • Promotion of model lightweighting and edge deployment, enabling RWKV models to operate efficiently in resource-constrained environments such as mobile devices and IoT devices.
  • Developing More ToC AI Applications:
    • Expansion into diverse application scenarios, applying RWKV technology to a wider range of consumer-level settings.
    • Emphasis on user feedback to continuously optimize product design and meet user needs.
  • Supporting Ecosystem Development:
    • Building a more vibrant developer community to lower the barrier to entry for using RWKV.
    • Organizing RWKV technology exchange activities and related competitions, such as the publicly announced "2025 RWKV Ecosystem Content Collection Competition," and planning to launch the "2025 RWKV Ecosystem Annual Awards" with defined award settings and judging rules.
    • Promoting industry collaboration with upstream and downstream companies to jointly advance the application and popularization of the RWKV architecture.
    • Actively collaborating with universities, research institutions, and open-source communities to promote the open-source development of RWKV technology.

RWKV-7: A New Power for Edge AI

YuanShi Intelligence's latest RWKV-7 architecture adopts a dynamic state evolution mechanism, disrupting traditional attention/linear attention models. It not only possesses strong contextual learning capabilities but also achieves true continuous learning. This means the model can continuously self-optimize and improve based on new data in practical applications, significantly enhancing its adaptability and efficiency.

RWKV-7 maintains 100% Recurrent Neural Network (RNN) characteristics while also possessing excellent long-text processing capabilities, easily handling complex text processing tasks. For example, the RWKV-7-World 0.1B model, after pre-training with a 4k context length, can perfectly pass the 16k context length "needle in a haystack" test without any fine-tuning.

RWKV Technology: Industry Recognition and Thriving Open-Source Ecosystem

Since its inception, the RWKV architecture has garnered widespread attention and application due to its efficiency and practicality, becoming one of the most notable technical solutions in the AI field. Notably, in September 2024, the RWKV community discovered that Microsoft Windows had integrated the RWKV runtime library into its Office component updates. This means that hundreds of millions of Windows devices worldwide are already equipped with RWKV technology, which is expected to support certain functions in Windows, such as local Copilot and local memory recallers. This fully demonstrates RWKV's advantages in edge deployment and low power consumption, as well as its enormous potential in practical applications.

The thriving open-source ecosystem of RWKV has also attracted the participation of numerous leading companies and research institutions. For example, companies like Alibaba, Tencent, and Horizon Robotics have conducted research on multi-modal information processing and embodied intelligence based on RWKV. Additionally, universities such as Zhejiang University and Southern University of Science and Technology have carried out various innovative studies based on RWKV, including multi-modal models, brain-like models, and decision-making models, further promoting the diversified application and breakthroughs in AI technology.

Currently, the RWKV official website has included over 40 papers written by various universities and companies on the use of RWKV, which fully demonstrates the feasibility and capability of RWKV in language, multi-modal, and time-series domains.

YuanShi Intelligence, as a high-tech enterprise focused on the research and development of large model architectures and AI applications, centers its core technology around the RWKV architecture. The company is committed to creating efficient and lightweight AI models to break through the bottlenecks of traditional Transformer architectures, achieving more efficient edge deployment and broader application scenarios.