benchgap
Vision & documents

OfficeQA leaderboard

As of 2026-11-18, the highest measured score on OfficeQA is 83.4% by Claude Fable 5.2. 9 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.283.4%measured
2Claude Fable 5.180.2%measured
3Claude Opus 5.578.9%measured
4Kimi K3.1 Pro78.6%estimated ± 0.2 pp, medium confidence
5DeepSeek V4.1 Flash 101778.5%estimated ± 2.6 pp, medium confidence
6Inkling 2 Small78.5%estimated ± 2.6 pp, low confidence
7Claude Mythos 578.4%estimated ± 2.7 pp, medium confidence
8Claude Opus 578.1%measured
9Claude Opus 4.877.4%estimated ± 0.2 pp, medium confidence
10Hy4 preview77.4%estimated ± 0.2 pp, medium confidence
11Claude Sonnet 5.576.9%measured
12Kimi K375.2%estimated ± 0.2 pp, low confidence
13GLM-5.3-Flash74.7%estimated ± 0.2 pp, low confidence
14Step 5 Preview73.9%estimated ± 0.2 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General