{"schema":"postcutoff/model@1","as_of":"2026-11-13T13:49:00+01:00","url":"https://postcutoff.com/m/clover-35b-a3b-preview/","md":"https://postcutoff.com/m/clover-35b-a3b-preview/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"clover-35b-a3b-preview","name":"Clover-35B-A3B-Preview","org":"DiffuSpace","family":"Clover","released":"2026-10-28","status":"preview","type":"reasoning-llm","modality_in":["text"],"modality_out":["text"],"open_weights":true,"model_license":"apache-2.0","context_window":131072,"max_output":null,"knowledge_cutoff":null,"pricing":null,"price_line":"Price not published","access":[{"provider":"Hugging Face (open weights, BF16 ~69 GB; custom code, trust_remote_code)","model_id":"Diffuspace/Clover-35B-A3B-Preview","url":"https://huggingface.co/Diffuspace/Clover-35B-A3B-Preview"},{"provider":"GitHub (SGLang and llama.cpp inference backends, self-speculative decoding)","url":"https://github.com/DiffuSpace/Clover-35B-A3B-Preview"},{"provider":"ModelScope","url":"https://modelscope.cn/models/Diffuspace/Clover-35B-A3B-Preview"},{"provider":"Web demo (application required)","url":"https://chat.diffuspace.ai/"}],"capabilities":[{"name":"Block-diffusion MoE language model for agents","detail":"Text diffusion model derived from Qwen3.6-35B-A3B (35B total, ~3B active, 256 experts top-8): generates 32-token blocks in parallel by unmasking confident positions; DiffuSpace reports SWE-bench Verified 73.6, Terminal-Bench 2.1 51.7 and τ³ 65.3, above Qwen3.6-35B-A3B and the diffusion models LLaDA2.2 flash and Mercury 2.5 (vendor numbers).","first":false,"discovered":"launch","source":"https://diffuspace.ai/blog/clover-35b/"},{"name":"Fast single-stream generation","detail":"Vendor-measured 680 tokens/s at batch size 1 in BF16 on one GB300 on internal agent workloads, and 515 tokens/s on one H200 vs 365 tokens/s for a DFlash2-accelerated Qwen of the same size.","first":false,"discovered":"launch","source":"https://diffuspace.ai/blog/clover-35b/"}],"entry":"2026-10-28-diffuspace-clover-35b-a3b-diffusion-lm","notes":"First model from DiffuSpace, a company started by members of HKU NLP's Dream diffusion-LM team. Preview release; thinking mode on by default; generation needs the repository's custom code (Transformers 5.19). Long generations can repeat, so the defaults add presence, repetition and length penalties. All benchmarks are the vendor's own; no independent leaderboard results yet.","verified":"2026-10-29","body_md":"Open-weight diffusion language models have usually trailed autoregressive ones of the same size on agentic coding; DiffuSpace\nclaims Clover beats its own autoregressive base (Qwen3.6-35B-A3B) on SWE-bench Verified. Quick start (model card): Transformers\n5.19, `trust_remote_code=True`, BF16 on one GPU with ~69 GB plus inference state, `generate(..., block_size=32, threshold=0.95)`.","page_url":"https://postcutoff.com/m/clover-35b-a3b-preview/","events_after":null,"major_after":null,"historic_after":null,"missing_at_launch":null,"events_since_release":333,"you_url":"https://postcutoff.com/you/clover-35b-a3b-preview/","briefings":null}