Chinese AI startup DeepSeek, based in Hangzhou, launched its V4.1 Flash model on September 10, a system with 552 billion parameters offered at a fraction of competitors’ prices. The model uses a mixture-of-experts architecture with native multimodal capabilities, optimized for coding and agentic workflows. The announcement triggered a negative market reaction: shares in MiniMax and Z.ai dropped more than 8%, while Alibaba fell over 2%. Prior DeepSeek models cost roughly 3 cents per test compared to $3.15 for Anthropic’s Claude Fable 5, a price difference exceeding a factor of 100. The company is now reportedly preparing an IPO on Shanghai’s STAR Market.
Source: Read the original article
