Databricks Launches OfficeQA Pro V2 Benchmark

Databricks has launched OfficeQA Pro V2, a new benchmark based on 120,000 PDF documents from the US Treasury to test AI agents' complex document processing capabilities, where current leading AI models average only 26% accuracy.

2026-08-07 ~ 2026-08-07 · 2 related posts

• OfficeQA Pro V2 Released with 120K Treasury PDFs, Frontier Agents Average 26% Accuracy — jefrankle · 2026-08-07
• New OfficeQA Pro V2 Benchmark: Frontier AI Agents Average Just 26% Accuracy — jefrankle · 2026-08-07