List your company

Spreadsheet Understanding and Manipulation on SpreadsheetBench Hard Restriction

66.67Cell-Level AccuracyHuman Performance
-2.666815.334133.33551.3359Apr 14, 2026

Evaluation results

MethodLinks
Human Performance2026.0466.675562
SpreadsheetAgentBackbone=Qwen3-Coder-480B2026.0436.931.0534.65
SpreadsheetAgentBackbone=GPT-OSS-120B2026.0432.829.3431.47
TreeThinkerBackbone=Qwen3-Coder-480B2026.0431.9127.9230.37
SpreadsheetAgentBackbone=GPT-OSS-20B2026.0427.8125.6429.25
SpreadsheetAgentBackbone=Qwen3-235B2026.0425.3125.0725.22
GPT-OSS-120BBackbone=GPT-OSS-120B2026.0424.9623.9324.56
TreeThinkerBackbone=GPT-OSS-120B2026.0424.4226.525.22
Qwen3-Coder-480BBackbone=Qwen3-Coder-480B2026.0422.8227.0724.45
GPT-OSS-20BBackbone=GPT-OSS-20B2026.0419.4321.3720.18
TreeThinkerBackbone=GPT-OSS-20B2026.0418.5422.5120.07
TreeThinkerBackbone=Qwen3-235B2026.0417.2923.6519.74
SpreadsheetAgentBackbone=Qwen3-30B2026.0416.0422.2218.42
Qwen3-235BBackbone=Qwen3-235B2026.0415.3322.5118.09
GPT-4oBackbone=GPT-4o2026.0410.5217.6613.27
TreeThinkerBackbone=Qwen3-30B2026.0410.1619.0913.6
Qwen3-30BBackbone=Qwen3-30B2026.049.4518.2312.83
GPT-3.5Backbone=GPT-3.52026.042.59.975.37
Mixtral-8x7BBackbone=Mixtral-8x7B2026.042.323.712.85
Llama-3-70BBackbone=Llama-3-70B2026.040.717.143.18
BinderBackbone=GPT-3.52026.04000