Spreadsheet Arena

At a glance

Spreadsheet Arena compares AI-generated spreadsheets through blind, head-to-head outputs and user preference votes. Its leaderboard aggregates those comparisons. This is a way to explore model fit for spreadsheet creation, rather than a contest in which people compete at spreadsheet skills.

Good fit

Worth checking when you want a shortlist of models to try on your own spreadsheet task. Battle counts, confidence intervals and rank ranges make the comparison more informative than a bare winner badge. Inspect those before treating a small score difference as meaningful.

Consider something else when

You need proof that a workbook’s formulas are correct or a private workspace for company data. Preference votes describe which output people chose, not an independent audit of calculations. The privacy notice says test prompts and usernames may be made public.

Start here

Start with Full leaderboard and compare the rank ranges and battle counts of nearby models. For participation, the published flow is to write a prompt, compare blind outputs and vote. Use a fictional task with checkable totals; verify formulas and assumptions before reusing any resulting workbook.

Cost and requirements

The rankings and public comparison entry point were accessible without an account or payment. A generation quota or ongoing price was not established by those pages. Generation uses third-party AI services; submitted prompts, outputs and votes are collected under the site’s policy.

Limits to know

Ratings use a Bradley-Terry variant on an Elo-like scale. At inspection, the first-ranked model’s displayed rank range was 1–4, showing why first place is not a definitive ordering. Community tasks and preferences need not match your workload. The terms permit broad service and product-improvement use of submitted content; keep sensitive spreadsheets out of an exploratory trial.

Still unclear

• What generation limits or charges apply to a new participant?