DiffuSpace Releases New Open-Source Clover-35B Model
The 35-billion parameter AI model utilizes a mixture-of-experts design to boost coding performance.
Updated on Oct. 29, 2026 in Artificial Intelligence
Live Poll
Do you trust open-source artificial intelligence tools to handle your professional coding or research tasks?
DiffuSpace has released its Clover-35B-A3B-Preview model as open-source software, featuring a mixture-of-experts architecture. The model is now accessible to developers via Hugging Face, GitHub, and ModelScope.
Why it matters
The release introduces a new architecture using block diffusion and a Gated DeltaNet backbone to improve sequence generation. This model provides high-speed inference performance for developers working with coding-focused AI tools.
The Clover-35B-A3B-Preview model contains 35 billion total parameters with 3 billion active parameters per token, achieving speeds of 515 tokens per second on an NVIDIA H200. It recorded a 65.3 score on the tau-cubed benchmark and 67.9 on BFCL-v4.
The players
DiffuSpace
DiffuSpace is a technology organization focused on developing advanced artificial intelligence models and coding tools.
Hugging Face
Hugging Face is a collaborative platform that hosts machine learning models and datasets for the developer community.
NVIDIA
NVIDIA is a global technology company known for designing graphics processing units and hardware platforms for artificial intelligence.
The details
Engineered with block diffusion and learnable block-size calibration, the model processes output blocks in sequence using parallel denoising. It is distributed under the Apache 2.0 license to facilitate community integration.
Timeline
DiffuSpace released the model on October 29, 2026.
The Tech Race
The release utilizes the Apache 2.0 license to foster developer collaboration, following a common industry practice for open-source AI models. This model competes in a market where developers prioritize efficiency and standardized licensing to integrate AI into existing software ecosystems.
Developers can now integrate this model into their workflows to improve coding tasks and sequence generation efficiency. The availability of the model on major repositories allows for immediate testing and implementation in local or cloud environments.
The takeaway
The introduction of block diffusion-based models marks a shift in how efficiently large-scale models handle sequential coding tasks. Developers should monitor the upcoming agent harness for further improvements in autonomous workflow capabilities.
Further reading
For more on the latest developments in the field, visit the Artificial Intelligence section.
Source note: This article includes information reported by Pandaily.
Live Poll
Do you trust open-source artificial intelligence tools to handle your professional coding or research tasks?
