---
id: "2026-10-28-diffuspace-clover-35b-a3b-diffusion-lm"
url: "https://postcutoff.com/e/2026-10-28-diffuspace-clover-35b-a3b-diffusion-lm/"
as_of: "2026-11-13T13:49:00+01:00"
date: "2026-10-28"
date_precision: day
category: open-source
importance: 2
confidence: high
status: [Confirmed]
sources: 4
editor: human editor
human_review: null
version: "2026-10-29"
---

As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-10-28-diffuspace-clover-35b-a3b-diffusion-lm/

# DiffuSpace open-sources Clover-35B-A3B-Preview, a block-diffusion MoE language model that it says beats its Qwen3.6 base on agentic coding (SWE-bench Verified 73.6)

DiffuSpace, founded by members of the University of Hong Kong NLP group behind the Dream diffusion models, released Clover-35B-A3B-Preview under Apache 2.0 around Oct 28, 2026. It converts Qwen3.6-35B-A3B into a text-diffusion model that writes 32-token blocks in parallel, with a 128K context. DiffuSpace reports SWE-bench Verified 73.6, Terminal-Bench 2.1 51.7 and 680 tokens/s at batch size 1 on one GB300, ahead of both its autoregressive base and other diffusion LMs (vendor numbers).

## Key facts
- Architecture (model card): Qwen3.6-35B-A3B backbone kept (hybrid Gated DeltaNet / full attention, 40 layers, 256 routed experts, top-8, one shared expert); bidirectional attention inside each block; block size 32, confidence threshold 0.95
- Vendor benchmarks (blog): SWE-bench Verified 73.6, Terminal-Bench 2.1 51.7, τ³ Overall 65.3 (best of the compared models on six agent tests, against LLaDA2.2 flash, Mercury 2.5, Qwen3.6-35B-A3B and Qwen3.5-35B-A3B); GPQA Diamond 83.8, ZebraLogic 95.8, AA-LCR 72.0
- Speed (vendor): 680 tokens/s overall on internal agent data, BF16, batch size 1, one NVIDIA GB300; 515 tokens/s on one H200 vs 365 for a DFlash2-accelerated Qwen model
- Release: Hugging Face repo created Oct 24, GitHub README and benchmark image added Oct 27–28; community GGUF and MLX quantizations appeared Oct 28; Pandaily covered it Oct 29
- Lineage: Dream 7B (HKU NLP, April 2025) has 2.7M+ downloads, per DiffuSpace

## What happened

A start-up formed by members of HKU NLP's Dream team turned an open autoregressive MoE model into a diffusion model and claims it is now both better and faster at
agentic coding than the model it came from. Weights, inference code for SGLang and llama.cpp, and a web demo are public.

## Why it matters

Diffusion text models promise much faster generation, which matters most for agents that loop many times per task. Until now open
diffusion LMs trailed autoregressive models of the same size; the claim still needs independent benchmark runs.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 181 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 120 days after its cutoff
- Gemini 4 Hydrogen (training cutoff March 2026): 211 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 150 days after its cutoff

## Sources

1. [DiffuSpace blog: Clover-35B-A3B-Preview (in Chinese)](https://diffuspace.ai/blog/clover-35b/) (diffuspace.ai, official)
2. [Hugging Face: Diffuspace/Clover-35B-A3B-Preview (model card)](https://huggingface.co/Diffuspace/Clover-35B-A3B-Preview) (huggingface.co, code)
3. [GitHub: DiffuSpace/Clover-35B-A3B-Preview](https://github.com/DiffuSpace/Clover-35B-A3B-Preview) (github.com, code)
4. [Pandaily: DiffuSpace open-sources Clover-35B-A3B (Oct 29)](https://pandaily.com/diffuspace-clover-35b-a3b-preview-open-source-diffusion-language-model-swe-bench-73-6) (pandaily.com, press)

## Changes

- 2026-10-29 (filed): Created (model card, blog and GitHub read directly; news agent)

## Related

- Models: [Clover-35B-A3B-Preview](https://postcutoff.com/m/clover-35b-a3b-preview/) (`clover-35b-a3b-preview`)