DeepSeek launches image recognition mode, adding AI vision capabilities

DeepSeek has introduced a new 'image recognition mode' on its chat interface, alongside existing 'expert' and 'flash' modes. The feature is initially available for beta testing to select users on the chatbot website and mobile app. The limited release follows the Hangzhou-based company's launch of its flagship model V4 and subsequent price cuts. Multimodal team leader Chen Xiaokang announced the update on social media on Wednesday. The addition addresses a key gap in DeepSeek's offerings, as its breakout in January 2025 was driven by powerful reasoning and cost-efficiency but lacked multimodal capabilities, which are increasingly seen as essential for moving beyond text conversations into more complex domains.