A new technical report introduces Clover-35B-A3B-Preview (Clover), a mixture-of-experts diffusion language model with 35 billion total parameters and 3 billion activated parameters. Clover utilizes Gated Delta Networks and achieves state-of-the-art performance on seven agent metrics, including high scores on SWE-bench Verified and Terminal-Bench 2.1. The model demonstrates impressive generation speeds of 680 tokens/s on a single NVIDIA GB300 GPU and includes a decision-model variant that enhances question accuracy. AI
IMPACT Sets new SOTA on agent benchmarks, potentially influencing future development in LLM-based agents and their deployment efficiency.
RANK_REASON The cluster contains a technical report detailing a new language model and its performance on benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- Clover
- Clover-35B-A3B-Preview
- Gated Delta Networks
- Hugging Face
- NVIDIA GB300
- SWE-bench Verified
- Terminal-Bench 2.1
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →