SpreadsheetBench
MiniMax-M3 89.3% leads since June 2026 MiniMax-M3 89.3% leads since June 2026 Real spreadsheet manipulation questions from Excel forums. Output workbook is checked cell-by-cell against reference solutions.
Kimi K3 reports 'SpreadsheetBench 2'. We did not find a separate public repo for v2.
- Released
- June 2024
- Kind
- Documents and spreadsheets
- Level
- University
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Strongest model tested
-
Qwen3.7-Max #49 on the capability index - Last in a launch post
- June 2026
- Last model added
- June 2026
- Scores refreshed
- November 12th, 2026
In the authors’ words
We introduce SpreadsheetBench, a challenging spreadsheet manipulation benchmark exclusively derived from real-world scenarios, designed to immerse current large language models (LLMs) in the actual workflow of spreadsheet users. Unlike existing benchmarks that rely on synthesized queries and simplified spreadsheet files, SpreadsheetBench is built from 912 real questions gathered from online Excel forums, which reflect the intricate needs of users. The associated spreadsheets from the forums contain a variety of tabular data such as multiple tables, non-standard relational tables, and abundant non-textual elements. Furthermore, we propose a more reliable evaluation metric akin to online judge platforms, where multiple spreadsheet files are created as test cases for each instruction, ensuring the evaluation of robust solutions capable of handling spreadsheets with varying values. Our comprehensive evaluation of various LLMs under both single-round and multi-round inference settings reveals a substantial gap between the state-of-the-art (SOTA) models and human performance, highlighting the benchmark's difficulty.
MiniMax-M3 leads with 89.3% among models released by June 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
ChatGPT agent leads with 35.3% across all independent runs
Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.
MiniMax-M3 leads with 89.3%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
The newest model with a score here came out in June 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 MiniMax-M3 89.3%
- 2 Qwen3.7-Max 87%
- 3 Qwen3.7-Plus 86.3%
- 4 Claude Opus 4.5 64.3%
- 5 ChatGPT agent 45.5%
- 6
ERNIE 5.0 40.1%
Source: Labs’ model cards
- 1 ChatGPT agent 35.3%
- 2 o3 23.3%
Source: Official SpreadsheetBench leaderboard
- 1 MiniMax-M3 89.3%
- 2 Qwen3.7-Max 87%
- 3 Qwen3.7-Plus 86.3%
- 4 Claude Opus 4.5 64.3%
- 5 ChatGPT agent 45.5%
- 6 ERNIE 5.0 40.1%
- 7 o3 23.3%
Sources: Labs’ model cards and official SpreadsheetBench leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | MiniMax-M3 | 89.3% | Jun 2026 | Lab model card | |
| 2 | Qwen3.7-Max | 87% | May 2026 | Lab model card | |
| 3 | Qwen3.7-Plus | 86.3% | Jun 2026 | Lab model card | |
| 4 | Claude Opus 4.5 | 64.3% | Nov 2025 | Lab model card | |
| 5 | ChatGPT agent | 45.5% | Jul 2025 | Lab model card | |
| 6 | ERNIE 5.0 | 40.1% | Feb 2026 | Lab model card |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | ChatGPT agent | 35.3% | Jul 2025 | Official leaderboard | |
| 2 | o3 | 23.3% | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | MiniMax-M3 | 89.3% | Jun 2026 | Lab model card | |
| 2 | Qwen3.7-Max | 87% | May 2026 | Lab model card | |
| 3 | Qwen3.7-Plus | 86.3% | Jun 2026 | Lab model card | |
| 4 | Claude Opus 4.5 | 64.3% | Nov 2025 | Lab model card | |
| 5 | ChatGPT agent | 45.5% | Jul 2025 | Lab model card | |
| 6 | ERNIE 5.0 | 40.1% | Feb 2026 | Lab model card | |
| 7 | o3 | 23.3% | Apr 2025 | Official leaderboard |
The benchmark's own leaderboard has 14 entries, 2 of them models we track. Switch the scores above to rank them together.
4 models have held number one in labs’ own numbers since its release in June 2024 Number one over time 5 models have held number one with claims included since its release in June 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Qwen’s own number for Qwen3.7-Max passed 85% in May 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official SpreadsheetBench leaderboard and labs’ model cards
2 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Jun 1, 2026 | Qwen3.7-Plus | 86.3 | SpreadSheetBench-v1 |
| Jun 1, 2026 | MiniMax-M3 | 89.3 | SpreadSheetBench-v1 |
| May 20, 2026 | Qwen3.7-Max | 87 | SpreadSheetBench-v1 |
| Feb 6, 2026 | ERNIE 5.0 | 40.1 | SpreadSheetBench |
| Jul 17, 2025 | ChatGPT agent | 45.5 | SpreadsheetBench, OSX LibreOffice, with .xlsx access 45.5 (without 35.3) |
2 system cards report it, from 2 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Feb 4, 2026 | ERNIE 5.0 technical report | 40.1 | SpreadSheetBench (Table 2) | Capabilities |
| Nov 24, 2025 | Claude Opus 4.5 system card | 64.3 | SpreadsheetBench, full 912 problems, no extended thinking, avg of 5 trials (2.15) | Capabilities |
One lab claim next to the leaderboard
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| ChatGPT agent | 45.5% SpreadsheetBench, OSX LibreOffice, with .xlsx access 45.5 (without 35.3) | 35.3% Leaderboard | +10 pts |
More in Economic work
- JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
- GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate.
Seed2.1 Pro 87.9% - GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
- AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
- CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%
- OfficeQA In use Grounded reasoning over a large corpus of U.S. Treasury Bulletin documents: locate tables and compute precise numeric answers. Claude Opus 4.7 86.3%