Provisional: 1 test run. Every score below comes from a single run of our own test. See how we test.
The test
Total sales by region from a small table containing a blank cell, then write formulas, without inventing a value for the blank.
Results
| Tool | Plan tested | Score |
|---|---|---|
| ChatGPT | Default "Instant" model | 91/100 Provisional: 1 test run |
| Claude | Opus, paid plan | 89/100 Provisional: 1 test run |
ChatGPT: 91/100 (Provisional: 1 test run)
Correct totals; no invented numbers; formulas (MAXIFS) work in Excel 2019+ and Google Sheets. Did not flag the blank in the third task.
| Criterion | Points |
|---|---|
| Accuracy | 30/30 |
| Data honesty | 16/20 |
| Clarity | 17/20 |
| Portability | 18/20 |
| Friction | 10/10 |
Claude: 89/100 (Provisional: 1 test run)
Correct totals and flagged the blank every time. Used XMATCH/LET and wrongly said they work in Google Sheets.
| Criterion | Points |
|---|---|
| Accuracy | 30/30 |
| Data honesty | 17/20 |
| Clarity | 19/20 |
| Portability | 13/20 |
| Friction | 10/10 |
Caveats
Both tools got every total right; the difference is mostly formula portability. We plan a second run of every tool before treating any ranking as settled.