Image & vision

ParseBench (LlamaIndex)

by LlamaIndex

What is ParseBench (LlamaIndex)? ParseBench (LlamaIndex) is an image & vision eval made by LlamaIndex. About 2,000 human-verified pages from insurance, finance and government documents, used to check whether PDF parsers keep what an agent needs: table structure, chart data, faithful content, meaningful formatting such as strikethrough, and visual grounding. Scoring is deterministic and rule-based by default. Built by LlamaIndex, whose LlamaParse is on its leaderboard.

Page last updated 2026-10-18 · Evals Review

Editor note

About 2,000 human-verified pages from insurance, finance and government documents, used to check whether PDF parsers keep what an agent needs: table structure, chart data, faithful content, meaningful formatting such as strikethrough, and visual grounding. Scoring is deterministic and rule-based by default. Built by LlamaIndex, whose LlamaParse is on its leaderboard.

What it tests
Document parsers and VLMs turning PDFs into structured output agents can act on.
Original source
www.parsebench.ai/
Last updated
Repo last commit 2026-10-20
License
Apache-2.0

Facts come only from the linked source (checked 15–16 Oct 2026). Credit belongs to the creators. We don't score or rank evals: how we review.

Reviews

No reviews yet.

Reviews are written by people who sign in with GitHub, one per account per eval. We can't confirm that each reviewer used the eval; we remove reviews that break our rules, never just for being negative, and we never add, buy or edit reviews. Averages, when shown, come only from these reviews.

Reviews from practitioners who sign in with GitHub open soon.

Related image & vision evals

  • MMMU — Expert-level multimodal understanding and reasoning.
  • MathVista — Mathematical reasoning over figures, plots and images.
  • MMBench — Broad perception and reasoning skills of vision-language models.
  • ChartQA — Question answering over charts.

All Image & vision evals · Comparison table