DeepSeek AI News
Latest news, updates, and announcements from DeepSeek AI.
Chinese regulators granted DeepSeek conditional access to Nvidia H200 chips, joining ByteDance and A...
Chinese regulators granted DeepSeek conditional access to Nvidia H200 chips, joining ByteDance and Alibaba in a pool cleared for over 400,000 units. This shift prioritizes firms capable of maximizing...
DeepSeek has emerged as a dark horse in the AI sector, securing a 4% global market share by offering...
DeepSeek has emerged as a dark horse in the AI sector, securing a 4% global market share by offering high-performance models at roughly 1/27th the cost of Western alternatives. This open-source strate...
DeepSeek has launched its V3.2 model series, reaching parity with GPT-5 and Gemini 3.0 Pro through a...
DeepSeek has launched its V3.2 model series, reaching parity with GPT-5 and Gemini 3.0 Pro through an open-weight framework. The 685B parameter Mixture-of-Experts system achieved gold-medal status in...
DeepSeek is set to launch its next-generation AI model, V4, in mid-February 2026 with a focus on eli...
DeepSeek is set to launch its next-generation AI model, V4, in mid-February 2026 with a focus on elite coding capabilities. Internal tests suggest V4 could surpass rivals like Anthropic Claude Opus 4....
DeepSeek released Math-V2, the first open-source AI model to achieve a gold medal score at the Inter...
DeepSeek released Math-V2, the first open-source AI model to achieve a gold medal score at the International Mathematical Olympiad. The model solved 5 of 6 IMO 2025 problems and scored 118/120 on the...
Chinese AI startup DeepSeek is reportedly developing a new AI model with banned Nvidia Blackwell chi...
Chinese AI startup DeepSeek is reportedly developing a new AI model with banned Nvidia Blackwell chips. These advanced semiconductors were allegedly smuggled into China through third countries, highli...
DeepSeek released V3.2 and V3.2-Speciale, 685B-parameter open-source models matching GPT-5 on reason...
DeepSeek released V3.2 and V3.2-Speciale, 685B-parameter open-source models matching GPT-5 on reasoning benchmarks and hitting IMO gold in math tasks. 🚀 Sparse attention cuts inference costs 50% with...
DeepSeek-Math-V2, an open-weights model, has set a new standard for AI reasoning, achieving a near-p...
DeepSeek-Math-V2, an open-weights model, has set a new standard for AI reasoning, achieving a near-perfect score of 118/120 on the Putnam 2024 competition. This landmark performance is powered by a no...
DeepSeek's new Math-V2 model fundamentally advances AI reasoning by introducing self-verification, m...
DeepSeek's new Math-V2 model fundamentally advances AI reasoning by introducing self-verification, moving beyond simple answer generation to reliable truth-finding. This 685B-parameter model achieved...