Home/New releases/Clover-35B-A3B-Preview
Clover-35B-A3B-Preview
An open-weight (Apache 2.0) diffusion language model that drafts text in parallel blocks; 35B total, about 3B active.
DiffuSpace published Clover-35B-A3B-Preview on October 27, 2026. It is a mixture-of-experts diffusion language model with 35B total parameters and about 3B active, a thinking mode, and a 128K context. This is a preview release. Weights are Apache 2.0 on Hugging Face and ModelScope. DiffuSpace says SGLang and llama.cpp inference are supported. The Hugging Face card says BF16 weights are about 69 GB before inference overhead. DiffuSpace's number: SWE-bench Verified 73.6. The team comes from the HKU NLP group behind Dream 7B. Last checked October 29, 2026.
Where it ranks
This is a dated release only. It does not have a rank on a use-case list. Feed label: Vibe coding · AI agents.
Pricing
Free
Apache 2.0 weights. The Hugging Face card says BF16 weights are about 69 GB before inference overhead.
Checked
Added October 29, 2026
Last verified October 29, 2026
Dated releases
DiffuSpace published Clover-35B-A3B-Preview, an Apache 2.0 diffusion language model with 35B total parameters and about 3B active.