同花顺
同花顺|Oct 20, 2025 10:28
[DeepSeek Team Open Sources New Model DeepSeek-OCR: Massive Text Compression with Few Visual Tokens] The DeepSeek-AI team has released the paper 'DeepSeek-OCR: Contexts Optical Compression,' proposing a new method to compress long text contexts using the visual modality. The Hugging Face page shows that the model has 3B parameters.
+5
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads