Skip to content

Benchmark

SpreadsheetBench 2 Visualization

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
7
Scores recorded
7
Sources
1
IQ weight
0.99

Score distribution

38.1%
Mean
41.7%
Median
13.8%
Std dev
25%
10th percentile
45.8%
90th percentile
16.7%
Min
16.7%62.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.99 for chat models, measured over 7 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 4.6scaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data62.5%
2GPT 5.2scaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data45.8%
3Gemini 3.1 Pro Previewscaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data41.7%
4Kimi K2.5scaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data41.7%
5DeepSeek V3.2scaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data33.3%
6Qwen3.5 397B A17Bscaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data25%
7MiniMax M2.5scaffold: swe-agent, date: 2026-03-26· SpreadsheetBench 2 leaderboard data16.7%

Data credits: SpreadsheetBench (Renmin University of China and contributors), spreadsheetbench.github.io.