Harpd Product Ranking Dataset
One record per approved product on Harpd Rank: rank position, Rank Points (promotional placement), category, verification state and per-record updatedAt.
Citation hub
Every dataset Harpd publishes is free to quote, redistribute and build on — including commercially — under CC BY 4.0. Attribution is the only condition. Pick a dataset below and copy the line.
Quick answer
Yes. All 8 Harpd datasets are published under CC BY 4.0 and can be quoted, redistributed and reused commercially as long as you attribute Harpd and link the licence. Each dataset below carries a ready-to-copy citation with the data date it refers to, the methodology behind the numbers, and the coverage the numbers apply to.
Data source: Harpd canonical dataset registryLast updated: Evidence registerDataset indexMachine-readable graph
The as of date is the dataset's own data date — it moves when the data changes, never on a rebuild. A dataset with no date shown has no machine-readable timestamp yet, and we would rather show nothing than guess.
One record per approved product on Harpd Rank: rank position, Rank Points (promotional placement), category, verification state and per-record updatedAt.
Immutable closed-period snapshots of Harpd Rank, used to derive movement, new entrants and category shifts in research reports.
Product entities in the Harpd dataset: name, slug, category, capabilities, verification state, profile URL.
Harpd categories with product counts and board URLs.
Deliberately excludes Rank Points and ranking position. Publishes only signals that cannot be bought: verification state, listing age, first-party referred visits, declared repository stars.
One record per model with input/output token price, the official provider pricing page it was read from, and the date it was verified.
Long-format benchmark records (model, task, metric, score, cost, timestamp) with an explicit isModeled flag separating measured runs from modeled estimates.
Machine-readable export of the Harpd Rank rules, boards, entry conditions and disclosure.
Prefer the most specific object you can. Link the product profile for a product claim (every record carries a canonical url), the methodology page for a rule, and the dataset for an aggregate. State the as of date alongside the figure — the boards move, and a dated claim stays verifiable after they do.
Publishing the boundary is part of being citable. These are statements we deliberately do not make, because the data behind them does not exist.
A switch decision requires a real-task quality gate on the specific workload. No dataset covers that yet.
We publish the method, not a result.
Rank Points are promotional placement bought with Credits. Harpd holds no quality measurement that would support an ordering by merit.
Harpd has no first-party measurement of third-party traffic or reviews, and does not republish vendor-supplied figures as facts.
The non-sponsored signals Harpd can verify are published in /data/rank-open-data.json instead.
Every citation on this page is generated from the canonical dataset registry in code, not written by hand. A dataset becomes citable only when it has a published endpoint, a methodology page, a stated update cadence and a stated coverage — a new endpoint cannot quietly become a source without all four. Full methodology →
Underlying data date: