SpreadsheetBench v2 - Template Cell Match: leaderboard

Metric: Cell-match score (%, Modif.) on the 97 template-completion tasks 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 December 2026. 8 models tracked.

Top models

#ModelScore
1Claude Opus 4.6 (Thinking)91.97
2Gemini 3.1 Pro (Preview)86.06
3GPT-5.2 (High)83.84
4DeepSeek V3.2 (Thinking)75.19
5Kimi K2.567.08
6GLM-566.6
7Qwen 3.5 397B A17B65.62
8MiniMax-M2.550.14

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