DeepSeek V4 Open Source Goes Viral: 1.6T Parameters, 1M Context, Domestic Compute Support
On April 24, Chinese AI firm DeepSeek launched and open‑sourced V4 series: V4‑Pro (1.6T parameters) and V4‑Flash (284B), with 1M‑token context. A key highlight is full optimization for Huawei Ascend chips, delivering faster inference at lower cost than Western peers. It went viral globally on April 27 as a “domestic LLM milestone.”
#DeepSeekV4 #AI
Previous article
new
Microsoft and OpenAI Revised Cooperation, Ending Exclusive Tie
Google to Invest Up to $40B in Anthropic, Forging Compute & Ecosystem Ties
DeepSeek V4 Open Source Goes Viral: 1.6T Parameters, 1M Context, Domestic Compute Support
OpenAI “Trial of the Century” Begins April 27; Musk vs. Altman
TSMC Accelerates 3nm AI Chip Expansion to Reshape Global Computing Supply Chain
Google Gemini 3.0 Launches with 1M Token Context & Top Multimodal Performance
OpenAI Launches Workspace Agents for Enterprise AI Deployment
Tesla Q1 Profit Up 56%, Musk Warns of Severe Global AI Chip Shortage
AI Agents Boom: OpenClaw Hits 136K Stars, Multi-Agent Collaboration Becomes Norm
Google Releases 8th-Gen TPU to Compete for Global AI Computing Market