SpreadsheetBench v2 - Debugging Cell Match: leaderboard

Metric: Cell-match score (%, Modif.) on the 100 debugging tasks (locating and repairing injected spreadsheet errors) under the SWE-agent-based scaffold (bash, view_xlsx and submit tools, at most 50 turns), thinking mode on, high reasoning effort for GPT-5.2 and Gemini 3.1 Pro. Source: arxiv.org. Saturation forecast: Around April 2027. 8 models tracked.

Top models

#ModelScore
1Claude Opus 4.6 (Thinking)50.38
2Gemini 3.1 Pro (Preview)41.86
3GPT-5.2 (High)39.07
4GLM-529.06
5DeepSeek V3.2 (Thinking)28.71
6Qwen 3.5 397B A17B24.06
7Kimi K2.519.48
8MiniMax-M2.516.76

Interactive version: theaggregate.ai/benchmark?slug=spreadsheetbench-v2-debugging-cell-match · How It Works · Data refreshed daily, snapshot 2026-09-29.