Analytics agents
SpreadsheetBench
Real spreadsheet QA plus multi-sheet editing, debugging, and visualization.
As of 13 August 2026. Held out. RUCKBReasoning SpreadsheetBench. No explicit data license. Forum-derived workbooks need provenance review.
SpreadsheetBench mixes QA, editing, debugging, and visualization on real workbooks. Verified-answer accuracy and workbook-state checks stay separate.
Anthropic's Fable 5 launch notes that Fable 5 beats Opus 4.8 on an internal spreadsheet suite. That is not this public pack. No official SpreadsheetBench table for Fable 5 or GPT-5.6 is cited.
- Scale
- Forum-derived workbooks; pin the verified subset at acquisition
- Provenance
- RUCKBReasoning SpreadsheetBench. No explicit data license found; underlying forum workbooks need provenance review.
- Contamination
- Medium. Public since 2024.
- Cannot establish
- Any commercial artifact until permission and workbook provenance clear.
Metrics, kept separate
- Verified-answer accuracy
- Workbook-state checks
- Edit precision
- Formula validity
No cited public figure is installed for this pack. Method, scale, and gates still apply. A missing chart is not a pass rate.
What the figures bound
The pack is in the catalog because spreadsheet work is economically material. It stays held out until permission and provenance clear.
Internal vendor spreadsheet suites are not SpreadsheetBench.
What this page does not conclude
- A commercial run of SpreadsheetBench
- Fable 5 or GPT-5.6 on this public pack
- Excel-product substitution
- An overall spreadsheet winner
Cited as calibration on these memos — not a score of those products.
Not attached to a software, advisory, or automation memo yet.