DeepSeek 推出 Sparse Attention AI 模型:中國邁向 AGI 一大步

#AGI 中國初創公司 DeepSeek 推出新一代 AI 模型 DeepSeek-V3.2-Exp,並引入創新技術「DeepSeek Sparse Attention(DSA)」,聲稱能在保持輸出質素的同時,大幅降低長文本運算成本。此實驗性版本被視為邁向次世代 AI 架構的中繼步驟,亦顯示其與中國晶片商合作的意向。API 價格更下調一半,進一步提升其商業競爭力。

Why it matters: DeepSeek’s DSA cuts costs while preserving performance — a key move in scaling long-context AI models.

The big picture: This release signals China’s growing presence in foundational AI research and infrastructure.

Full article https://www.bloomberg.com/news/articles/2025-09-29/deepseek-debuts-sparse-attention-method-in-next-gen-ai-model
📌 一杯咖啡價錢連接 Web3 世界 https://patreon.com/wanszezit

Comments are closed.

Proudly powered by WordPress | Theme: Baskerville 2 by Anders Noren.

Up ↑