Data analysis
Assistants and spreadsheet copilots that analyse your data by writing and running code for you, the models behind them, and data agents and notebook tools you can run yourself.
Models are ranked by SpreadsheetBench 2 (models in SWE-agent scaffold). Tools with their own result on SpreadsheetBench (V1 full, 912 tasks) or SpreadsheetBench 2 (products and agents) are ranked by it (each board ranks one kind of product); other tools inherit the tier of the best model they offer. How we rank.
- Best overallBGemini 3.1 Pro (Preview)23.7% on SpreadsheetBench 2 (models in SWE-agent scaffold)
- Best valueNo qualifying option yet
- Best free startAWPS Office (WPS AI)Free tier · paid plans · 73.5% on SpreadsheetBench (V1 full, 912 tasks)
- Best local (≤24 GB GPU)No qualifying option yet
- Tier: how close to the best result on the task's main public benchmark
- Outlined: a tool inherits the tier of the best model it offers
- No public result yet
- Cloud · Local: runs on your own hardware
Models
The engines themselves. Use them through an API or inside the apps on the right.
Local
Nothing tracked yet
Apps
Ready-made apps with a UI: sign in or install, then start.
Cloud
A
WPS Office (WPS AI)73.5% on SpreadsheetBench (V1 full, 912 tasks)B
Gemini in Google Workspace70.5% on SpreadsheetBench (V1 full, 912 tasks)B
Geminivia Gemini 3.1 Pro (Preview)$C
Microsoft 365 Copilot57.2% on SpreadsheetBench (V1 full, 912 tasks)C
ChatGPT45.5% on SpreadsheetBench (V1 full, 912 tasks)C
Claude42.9% on SpreadsheetBench (V1 full, 912 tasks)–
Better Analyst (formerly Formula Bot)No public result yet–
Databricks GenieNo public result yet–
Gemini Notebook (formerly NotebookLM)No public result yet–
GPT for Work (becoming Autosheet)No public result yet–
HexNo public result yet–
JuliusNo public result yet–
Powerdrill BloomNo public result yet–
QuadraticNo public result yet–
RowsNo public result yet–
Snowflake CoWorkNo public result yet–
ThoughtSpotNo public result yet
WPS Office (WPS AI)73.5% on SpreadsheetBench (V1 full, 912 tasks)B Gemini in Google Workspace70.5% on SpreadsheetBench (V1 full, 912 tasks)B Geminivia Gemini 3.1 Pro (Preview)$C Microsoft 365 Copilot57.2% on SpreadsheetBench (V1 full, 912 tasks)C ChatGPT45.5% on SpreadsheetBench (V1 full, 912 tasks)C Claude42.9% on SpreadsheetBench (V1 full, 912 tasks)– Better Analyst (formerly Formula Bot)No public result yet– Databricks GenieNo public result yet– Gemini Notebook (formerly NotebookLM)No public result yet– GPT for Work (becoming Autosheet)No public result yet– HexNo public result yet– JuliusNo public result yet– Powerdrill BloomNo public result yet– QuadraticNo public result yet– RowsNo public result yet– Snowflake CoWorkNo public result yet– ThoughtSpotNo public result yetLocal
Nothing tracked yet
Data agents & notebooks
Agents and notebook tools that write and run analysis code on your data.
Cloud
Quality vs. setup
The same tools as the map, placed by how good the results can be (up) and how much setup they take (right). Top-left is the easiest way to top quality; further right buys control, privacy or price at the cost of setup. Models are left out: you reach them through these tools.
- Cloud
- Local: runs on your hardware
- Hollow: tier inherited from the best model the tool offers
Not placed (no public result yet): marimo, Jupyter AI, Deepnote, arito, Better Analyst (formerly Formula Bot), Databricks Genie, Decide, Fabi.ai, Gemini Notebook (formerly NotebookLM), Google Colab, GPT for Work (becoming Autosheet), Hex, Julius, Powerdrill Bloom, Quadratic, Rows, Snowflake CoWork, ThoughtSpot
Everything, in order
| Tier | Name | Approach | Runs | Rated by | Price | Skill |
|---|---|---|---|---|---|---|
| A | WPS Office (WPS AI) Kingsoft Office | Apps | Cloud | 73.5% on SpreadsheetBench (V1 full, 912 tasks) | Free tier | beginner |
| B |  Gemini 3.1 Pro (Preview)Google | Models | Cloud | 23.7% on SpreadsheetBench 2 (models in SWE-agent scaffold) | $4.5/M tokens | — |
| B | Gemini in Google Workspace Google | Apps | Cloud | 70.5% on SpreadsheetBench (V1 full, 912 tasks) | — | beginner |
| B | Gemini Google | Apps | Cloud | via Gemini 3.1 Pro (Preview) | Free tier · $4.99/mo | beginner |
| C |  ShortcutShortcut | Data agents & notebooks | Cloud | 59.3% on SpreadsheetBench (V1 full, 912 tasks) | Free tier · $100/mo | beginner |
| C | Microsoft 365 Copilot Microsoft | Apps | Cloud | 57.2% on SpreadsheetBench (V1 full, 912 tasks) | $25/mo | beginner |
| C | ChatGPT OpenAI | Apps | Cloud | 45.5% on SpreadsheetBench (V1 full, 912 tasks) | Free tier · $20/mo | beginner |
| C | Claude Anthropic | Apps | Cloud | 42.9% on SpreadsheetBench (V1 full, 912 tasks) | Free tier · $20/mo | beginner |
| – |  marimomarimo | Data agents & notebooks | Local | No public result yet | Free | intermediate |
| – |  Jupyter AIProject Jupyter | Data agents & notebooks | Local | No public result yet | Free | advanced |
| – |  DeepnoteDeepnote | Data agents & notebooks | Cloud | No public result yet | Free tier | intermediate |
| – |  aritoarito | Data agents & notebooks | Cloud | No public result yet | — | beginner |
| – | Better Analyst (formerly Formula Bot) Better Analyst | Apps | Cloud | No public result yet | Free tier | beginner |
| – | Databricks Genie Databricks | Apps | Cloud | No public result yet | — | beginner |
| – |  DecideDecide AI | Data agents & notebooks | Cloud | No public result yet | $7/mo | beginner |
| – |  Fabi.aiFabi.ai | Data agents & notebooks | Cloud | No public result yet | — | intermediate |
| – | Gemini Notebook (formerly NotebookLM) Google | Apps | Cloud | No public result yet | Free tier | beginner |
| – |  Google ColabGoogle | Data agents & notebooks | Cloud | No public result yet | Free tier | intermediate |
| – | GPT for Work (becoming Autosheet) Talarian | Apps | Cloud | No public result yet | Free tier · $25/mo | beginner |
| – | Hex Hex | Data agents & notebooks + Apps | Cloud | No public result yet | Free tier · $36/mo | intermediate |
| – | Julius Julius AI | Apps | Cloud | No public result yet | — | beginner |
| – | Powerdrill Bloom Powerdrill | Apps | Cloud | No public result yet | Free tier · $17/mo | beginner |
| – | Quadratic Quadratic | Apps | Cloud | No public result yet | Free tier · $18/mo | beginner |
| – | Rows Rows (part of Superhuman) | Apps | Cloud | No public result yet | Free tier · $8/mo | beginner |
| – | Snowflake CoWork Snowflake | Apps | Cloud | No public result yet | — | beginner |
| – | ThoughtSpot ThoughtSpot | Apps | Cloud | No public result yet | $25/mo | beginner |
Benchmarks used
- SpreadsheetBench 2 (models in SWE-agent scaffold) by SpreadsheetBench team (Renmin University of China, Aptura.AI, AfterQuery, Shortcut.AI) (automated tests, Overall accuracy on 321 end-to-end business spreadsheet tasks (financial modeling, templates, debugging, visualization)) · snapshot 19 Oct 2026Model rows of https://spreadsheetbench.github.io/data/leaderboard-v2-full.json (unchanged 2026-10-19): 8 models run by the authors in the same SWE-agent scaffold, all dated 2026-03-26 (the paper's runs; no model row added in the seven months since, although the product side of the board and the V1 boards keep receiving weekly submissions); the 3 product rows are in spreadsheetbench-2-tools. Visualization tasks are scored by a VLM judge against assertion checklists. Only Gemini 3.1 Pro is a current version of a tracked family; the leader (Claude Opus 4.6) is an older version, so the board's top entry is untracked and current models rank against it. Shortcut.AI is one of the V2 authors (it has no V2 row). No other public board ranks current frontier models on spreadsheet work (DABstep and Spider 2.0 rank self-run research and data agents). Re-checked 2026-10-19: leaderboard-v2-full.json unchanged (11 rows; model rows still dated 2026-03-26); site repository last commit 2026-10-11. Gemini 3.1 Pro remains the only current tracked frontier version listed.
- SpreadsheetBench (V1 full, 912 tasks) by SpreadsheetBench team (Renmin University of China, Tsinghua University, Zhipu AI) (automated tests, Overall Pass@1 across 912 real-world spreadsheet questions and 2,729 test spreadsheets, ranks tools) · snapshot 19 Oct 2026Rows from https://spreadsheetbench.github.io/data/leaderboard-v1-full.json (unchanged 2026-10-19): 13 products/agents plus one bare model (OpenAI o3, 23.3, an old version; no models record made). Newest entry 2026-08-30 (Catalyst); submissions are processed weekly and the site repository was last updated 2026-10-11. Unverified rows (Copilot in Excel Agent Mode, ChatGPT Agent, Claude Files Opus 4.1, o3; dated 2025-07 to 2025-09) were evaluated by third parties such as OpenAI and Microsoft; verified rows are run by the board through vendor APIs. Tracked: Gemini in Google Sheets, Shortcut.ai, Copilot in Excel, ChatGPT Agent, Claude Files, WPS AI.
- SpreadsheetBench 2 (products and agents) by SpreadsheetBench team (Renmin University of China, Aptura.AI, AfterQuery, Shortcut.AI) (automated tests, Overall accuracy on 321 end-to-end business spreadsheet tasks (financial modeling, templates, debugging, visualization), ranks tools) · snapshot 19 Oct 2026Product rows of the V2 board, all marked verified (unchanged 2026-10-19): arito (54.51, 2026-09-13), WPS AI (Kingsoft Office, 50.86, 2026-08-29), Decide Agent 2.0 (33.81, 2026-08-25). Submissions processed weekly; site repository last updated 2026-10-11. Tracked: arito, WPS AI, Decide. Re-checked 2026-10-19: unchanged.