Benchmarks
Measure the whole intelligence system.
Purveyors benchmarks compare controlled models and complete agent systems on the evidence, judgment, calibration, and operational constraints that shape real coffee decisions.
Available evaluation
CoffeeBench V1
20 cases · 400 subject trials
Coffee supply-chain analyst arena
A matched evaluation of one model across four system treatments. Harnessed systems beat Raw, but Purveyors did not clearly beat Pi and Parchment added no measured pairwise lift. Read the hypothesis audit, full pairwise matrix, failure modes, and missing data.
Includes
- All 1,800 pairwise ballots
- Findings and product implications
- Methodology and immutable digests
Explore CoffeeBench →
Purveyors
Coffee intelligence platform. Daily-normalized data from 40+ US specialty importers, turned into procurement-ready analytics for green coffee buyers and roasting teams.
© 2026 Purveyors.