DeepSeek image recognition mode begins gray-scale testing.

DeepSeek's multimodal "image recognition mode" has entered gray-scale testing, and some users can see the new entry on the homepage. After uploading an image, DeepSeek can understand the objects and scenes in the image like a human. Real-world testing shows that the image recognition mode can accurately identify animal species, recognize the precise coordinates of Lingyin Temple in Hangzhou through architectural and textual clues, and even detect visual trap images.