DeepSeek releases DeepSeek-OCR 2, enabling AI to "see" an image in the same logical order as humans.

DeepSeek has released a new DeepSeek-OCR 2 model, employing the innovative DeepEncoder V2 method. This allows AI to dynamically rearrange the parts of an image based on its meaning, rather than simply scanning from left to right mechanically. This approach simulates the logical flow that humans follow when viewing a scene. Ultimately, this model outperforms traditional vision-language models when processing images with complex layouts (such as documents or charts), achieving more intelligent and causal reasoning-driven visual understanding.