BREAKING
Ai2 Releases olmOCR 2
0
Bench score
0
pt
gain via RL
0
B
VLM params
Single VLM Pass to Markdown
1
PDF / PNG / JPEG
↓
2
7B VLM pass
↓
3
Markdown output
↓
4
Tables & LaTeX
Subcategory Accuracy
Tables
84.9
Multi-column
83.7
Old equations
82.3
Open and Low-Cost
olmOCR 2
Apache 2.0
●
Under $200 / 1M pages
●
Qwen2.5-VL based
●
Docker, vLLM, SGLang
GPT-4o cost
●
Roughly 32x higher
●
Closed pipeline
First Mile for RAG Pipelines
AI NEWS BLITZ
Ai2 has released olmOCR 2, an open-source toolkit that turns PDFs into clean Markdown.