# Introducing ExtractBench: The Most Comprehensive Benchmark…

Canonical: https://brew.new/browse/templates/email/pt1_k97zam6zygwyre4kwwgf1az6ds8e46a7

Brand: llamaindex.ai
Category: newsletter

![Preview of Introducing ExtractBench: The Most Comprehensive Benchmark…](https://cdn.brew.new/email-preview-78af85d9d8b5f194-tracking_r57ztm0gkhh3shkkrf5mczek918dtybb-1789012196736.png)

## Email content

Header_Newsletter_1

ExtractBench — Additional asset (1)

Hi John,

We just open-sourced ExtractBench, the most comprehensive benchmark for information extraction from enterprise documents.

Read the blog → https://www.llamaindex.ai/blog/introducing-extractbench

Our applied research team built it because we couldn't find a benchmark that measured what our customers ask about. It tests:

14 frontier systems tested on 370 real enterprise documents, 4,869 pages, 8 domains, 67 document types

Scored on value accuracy, visual grounding, and cost per page, with every document tagged so a low score tells you what failed

Agentic Plus, a new Extract tier also just launched on LlamaParse today. It leads the board at 95.6% and holds 94.4% on documents past fifty pages, at 8.1¢ a page.

The scientific paper, dataset, eval code, and leaderboard are fully open and reproducible. Our announcement post provides quick access to it all.

— The LlamaIndex Team

LlamaIndex, 405 Howard Street, San Francisco, California 94105, United States

Unsubscribe

Manage preferences

[Open and remix this design](https://brew.new/browse/templates/email/pt1_k97zam6zygwyre4kwwgf1az6ds8e46a7)

[Browse email designs](https://brew.new/browse/templates)