PulseAugur
EN
LIVE 09:19:09

DiffuSpace open-sources Clover-35B-A3B diffusion model for agents and coding

DiffuSpace has released Clover-35B-A3B-Preview, a 35-billion parameter Mixture-of-Experts diffusion language model. This model is designed for agent tasks and coding, achieving a score of 73.6 on SWE-bench Verified and processing 515 tokens per second on agent tasks using an H200 GPU. AI

IMPACT This release provides a new open-source model for agent and coding tasks, potentially advancing research and development in these areas.

RANK_REASON The cluster reports on the open-sourcing of a new language model with specific performance metrics. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Pandaily →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

DiffuSpace open-sources Clover-35B-A3B diffusion model for agents and coding

How we ranked this

Signal score
30 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster reports on the open-sourcing of a new language model with specific performance metrics. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Pandaily TIER_1 English(EN) · [email protected] (Pandaily) ·

    DiffuSpace Open-Sources Clover-35B-A3B, a Diffusion Language Model Built for Agents and Coding

    DiffuSpace has open-sourced Clover-35B-A3B-Preview, a 35B-parameter MoE diffusion language model. The team reports 73.6 on SWE-bench Verified and 515 tokens/s on one H200 in agent tasks.