# OfficeQA Pro ![Category](https://img.shields.io/badge/Category-Document%20QA-1f6feb) ![Metric](https://img.shields.io/badge/Metric-%25%20Accuracy-2da44e) ![Saturation](https://img.shields.io/badge/Saturation-Low-d29922) ![Created](https://img.shields.io/badge/Created-2026-555) > **Quick answer:** OfficeQA Pro tests a model's ability to answer questions grounded in real office documents — Word documents, Excel spreadsheets, PowerPoint presentations, and PDFs — requiring accurate extraction and reasoning over structured and semi-structured office content. Kimi K3 scores 63.3% as of July 2026. ## At a Glance **What it tests:** Question-answering accuracy over diverse office document formats, requiring both content extraction and reasoning across tables, slides, and formatted text. **Why it matters:** Office documents are a dominant real-world data source for enterprise AI assistants. OfficeQA Pro targets this practical, high-value use case directly. **Known limitations:** As an emerging benchmark, exact document sourcing and question design methodology are not yet independently published outside its citation by Moonshot AI. ## What OfficeQA Pro Measures OfficeQA Pro evaluates a model's ability to answer questions grounded in real office documents across common formats — Word, Excel, PowerPoint, and PDF. Questions require extracting and reasoning over structured elements (tables, cell references, slide content) as well as unstructured prose, testing practical document-understanding skills relevant to enterprise assistant use cases. ## Benchmark Specifications | Field | Value | |---|---| | Task category | Reasoning / document QA | | Metric | % accuracy | | Saturation | Low | | Created by | Not yet independently documented | ## How OfficeQA Pro Is Scored Models answer questions about the content of provided office documents, scored on % accuracy against ground-truth answers. ## State-of-the-Art Results | Rank | Model | Score | Source | Date | |---|---|---|---|---| | 1 | [Kimi K3](/models/moonshot-ai/kimi-k3) | 63.3% | [Kimi K3 technical report](https://www.kimi.com/blog/kimi-k3) | 2026-07 | *Score sourced from Moonshot AI's Kimi K3 announcement, July 2026.* ## OfficeQA Pro on Benchgen > No Benchgen results yet — [be the first to run OfficeQA Pro](/benchmarks/databricks/officeqa-pro). ## OfficeQA Pro vs Other Benchmarks | Benchmark | What it tests | Saturation | |---|---|---| | **OfficeQA Pro** | Office document question-answering | Low | | [SpreadsheetBench 2](/benchmarks/zhu-et-al/spreadsheetbench-2) | Spreadsheet-specific task completion | Low | | [OmniDocBench](/benchmarks/opendatalab/omnidocbench) | Diverse PDF document parsing | Low | | [BIRD-SQL Dev](/benchmarks/bird-bench/bird-sql-dev) | Structured data question-answering (SQL) | Medium | ## Run OfficeQA Pro on Your Model Benchgen lets you run OfficeQA Pro against your own model, tracking document QA accuracy across Word, Excel, PowerPoint, and PDF formats. ## Frequently Asked Questions

What is OfficeQA Pro?

OfficeQA Pro is a benchmark testing question-answering accuracy over real office documents — Word, Excel, PowerPoint, and PDF files.

What does a good score look like on OfficeQA Pro?

Kimi K3 reports 63.3% as of July 2026, a solid result reflecting practical document comprehension across common office formats.

Who created OfficeQA Pro?

OfficeQA Pro's originating team is not yet independently documented outside of its citation in Kimi K3's July 2026 technical report.

Rank	Model	Score
1	kimi-k3-1-pro	67.4
2	hy4-preview	66.2
3	kimi-k3	63.3

OfficeQA Pro — Results

OfficeQA Pro — Phases

OfficeQA Pro