Turn emails into revenue with Brew. No credit card, free credits to try.
llamaindex.ai · newsletter
Explore this email design and adapt it to your own brand. Review the copy, links, and offer before sending.
1920x1080
Hi there,
We're open sourcing ParseBench, the first document OCR benchmark built for the agentic era. Before an agent can approve a claim or extract terms from a contract, it needs to read the document correctly. ParseBench measures whether parsers actually do, across ~2,000 human-verified enterprise pages and 167,000+ test rules spanning tables, charts, content faithfulness, semantic formatting, and visual grounding.
👉 Read the full blog here for quick links to the HF dataset, github code, and arXiv paper.
We tested 14 methods. A few findings stood out:
Layout understanding still trips up even frontier VLMs, formatting scores range from 1% to 85%
Legacy IDP is decent at text extraction, but 1 in 10 pages still has meaningful errors
Charts are the great divider, most specialized parsers score below 6%
LlamaParse Agentic was the only method competitive across all five dimensions at 84.9% overall
If AI agents will truly automate knowledge work, they have to read our documents. Everything is open source. We'd love you to try it!
Best regards,
The LlamaIndex Team
Calling all enterprises: are you building with LlamaIndex? We are working hard to make LlamaIndex even more Enterprise-ready and have sneak peeks at our upcoming products available for partners. Interested? Get in touch.
Join our Discord community for live discussions and support.
Check out our GitHub for the latest code releases and documentation.
Outline_Text_Newsletter
LlamaIndex-logo
LlamaIndex, 405 Howard Street, San Francisco, California 94105, United States
Unsubscribe
Manage preferences
github
discord
X
YouTube
Browse more email designs