Xiaokang Chen, a researcher specializing in multimodal technology at DeepSeek, has officially announced the rollout of DeepSeek's image recognition mode across both its web and mobile App platforms. According to testing carried out by IThome, the App version of DeepSeek's image recognition mode still displays the message, 'The Image Understanding Feature is Currently in Beta Testing.' However, this notice is absent from the web version. The image recognition mode is now available alongside Quick Mode and Expert Mode options. Once enabled, users can directly upload images for DeepSeek to analyze and interpret, offering a functionality that goes well beyond mere text extraction. Moreover, earlier this April, DeepSeek also released technical insights into the multimodal model that powers the image recognition mode, introducing a core framework known as 'Thinking in Visual Primitives.'