ainewsblitz.com

Breaking

DeepSeek Releases DeepSeek-OCR 2 With 'Visual Causal Flow' for Layout-Aware Document Parsing

DeepSeek Releases DeepSeek-OCR 2 With 'Visual Causal Flow' for Layout-Aware Document Parsing
  • Foundation Models
  • Open Source
  • Media Generation

DeepSeek AI has open-sourced DeepSeek-OCR 2, a 3-billion-parameter vision-language model that converts images and PDFs into Markdown by reordering visual tokens to mimic how humans read a page. Released in late January 2026 under an Apache-2.0 license, the model pairs a redesigned encoder with a technique the team calls Visual Causal Flow, and it is available now on GitHub and Hugging Face, alongside a research paper.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,897 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year