Unlimited OCR Gen 2 Runs 5.6x Faster on Long Documents with Far Less VRAM
Wenxin's Unlimited OCR Gen 2 processes 100-page documents 5.6x faster, improves multi-page scores from 61.89 to 84.59, and slashes cache VRAM from 47GB to 1.77GB by keeping only the active segment in memory, mimicking how humans transcribe text.
2026-10-27 ~ 2026-10-27 · 4 related posts
- Episode 1: Baidu Open-Sources Unlimited OCR 2 for Thousand-Page Document Parsing(2026-10-26, 5 posts)
- Episode 2: Unlimited OCR Gen 2 Runs 5.6x Faster on Long Documents with Far Less VRAM(2026-10-27, 4 posts)
- Unlimited OCR Gen 2 Works Like a Human Copyist: Only the Active Paragraph Stays Hot — xiaohu · 2026-10-27
- Unlimited OCR Gen 2 Scores 84.59 on Multi-Page Bench, Up 23 Points — xiaohu · 2026-10-27
- Unlimited OCR Cuts Thick-Document KV Cache from 47GB to 1.77GB — xiaohu · 2026-10-27
- Unlimited OCR Gen 2 Generates 5.6x Faster on 100-Page Documents — xiaohu · 2026-10-27