{"slug": "deepseek-v3-2-exp", "title": "DeepSeek V3.2 Exp", "summary": "DeepSeek released V3.2 Exp, a 685B-parameter mixture-of-experts model with 37B active parameters per token, featuring DeepSeek Sparse Attention (DSA) for fine-grained sparse attention. The model, built on V3.1-Terminus and released under MIT license, was later superseded by V4 within about seven months.", "body_md": "# DeepSeek V3.2 Exp\n\nMoE workstation685B total, ~37B active per token (MoE, 256 routed experts) - shares the V3 backbone. Adds DeepSeek Sparse Attention (DSA): a lightning indexer (FP8, Hadamard dot-product) selects top-2048 tokens per query, then MLA runs only over those - the first fine-grained sparse attention from DeepSeek and the precursor to V4's CSA. Built on V3.1-Terminus. Experimental; superseded by V4 within ~7 months. Open weights under MIT.\n\ncoding\nreasoning\n\n- 685.0B\n- 128k\n- mit\n- Sep 2025\n\n## Scores\n\nCoding\n\n80\n\nReasoning\n\n86\n\nGeneral\n\n83\n\nPRICE HISTORY\n\n## Inference cost over time\n\nData accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API.\n\nLoading price history...", "url": "https://wpnews.pro/news/deepseek-v3-2-exp", "canonical_source": "https://tokenstead.ai/models/deepseek-v3-2-exp", "published_at": "2026-07-09 14:12:02+00:00", "updated_at": "2026-07-09 15:09:15.882342+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-research", "ai-products"], "entities": ["DeepSeek"], "alternates": {"html": "https://wpnews.pro/news/deepseek-v3-2-exp", "markdown": "https://wpnews.pro/news/deepseek-v3-2-exp.md", "text": "https://wpnews.pro/news/deepseek-v3-2-exp.txt", "jsonld": "https://wpnews.pro/news/deepseek-v3-2-exp.jsonld"}}