DiffuSpace, a Shenzhen-based diffusion language model startup, has completed two funding rounds with total financing close to 500 million yuan. Matrix Partners China, Shunwei Capital and Legend Capital co-led the deal, while Huawei Hubble and Horizon participated as follow-on investors. Several media outlets described it as the largest financing seen so far in the global diffusion language model segment. The company was founded in May this year by University of Hong Kong professor Kong Lingpeng and his PhD students Gong Shansan and Ye Jiacheng. The team had previously worked with Huawei’s Noah’s Ark Lab on Dream 7B and has already open-sourced the model weights. DiffuSpace is building on diffusion large language model, or dLLM, technology, which differs from the sequential text generation approach commonly used by models such as GPT and Claude. The startup is now training a new 30 billion-parameter model and plans to release it as open source in the near term, while focusing on code agents and local deployment across phones, cars and robots.
DiffuSpace, a Shenzhen-based diffusion language model company, has completed two funding rounds with total financing close to 500 million yuan. Matrix Partners China, Shunwei Capital and Legend Capital co-led the rounds, with Huawei Hubble and Horizon among the follow-on investors. Several media outlets said the deal is the largest financing in the global diffusion language model field.
DiffuSpace was founded in May this year by Kong Lingpeng, a professor at the University of Hong Kong, together with his PhD students Gong Shansan and Ye Jiacheng. The team previously worked with Huawei Noah’s Ark Lab to launch Dream 7B and has already open-sourced the model weights.
The company uses diffusion large language model, or dLLM, technology. Mainstream models such as GPT and Claude usually generate text in sequence, while diffusion models can handle content at multiple positions at the same time before gradually filling in and revising the output.
DiffuSpace is training a new 30 billion-parameter model and plans to release it and open-source it in the near term. The company said its focus will be on code agents and local deployment on devices including phones, cars and robots.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
