What Dedupe.io is
Dedupe.io, from the team behind the open-source dedupe Python library, does
probabilistic record linkage. You upload data, label a training set of "same or different" pairs, and
the model learns which field combinations predict a match, then scores every candidate pair. That is
real entity resolution: it can link records that share no exact field at all, and it reports
confidence rather than a yes or no.
Where the approaches split
The model costs setup: labeling sessions, and results that change with training. The threshold costs recall on hard cases: our fuzzy matching catches formatting variants and typos ("Smith, Jane" vs "jane smith" vs "Jane Smyth") but will never link "Bob" to "Robert", and we say so in the docs rather than implying otherwise. If your duplicates need inference, use an inference tool, theirs or AWS Entity Resolution. If your duplicates are formatting noise around a real key, which is most CRM exports and mailing lists, a deterministic pass is done before a model finishes training, and behaves the same way tomorrow.
How Dedupe.dev works, for contrast
Two halves. The browser tools are free and unmetered: your file is parsed and deduplicated by JavaScript in the tab, nothing is uploaded, and you can watch the empty Network panel while it runs. The HTTP API is for automation: exact or fuzzy matching, metered in rows (1,000 a month free with a key, 100,000 on Starter at $19/mo, 300,000 on Pro at $49/mo), with fuzzy requests up to 50,000 rows each. Duplicates are removed, never merged: the first occurrence survives, and re-importing the cleaned file is your step.
FAQ
Frequently asked questions
Is Dedupe.io the same as Dedupe.dev?
No relation. Dedupe.io is a machine-learning entity resolution service built on the dedupe Python library: you label example pairs, it trains a model, and it clusters likely matches with confidence scores. Dedupe.dev is a deterministic tool: exact matching, plus normalization-and-edit-distance fuzzy matching with a threshold you set.
When is a trained model worth it?
When your duplicates are genuinely ambiguous at scale: the same person as J. Smith, Smith John, and Jon Smith across databases with no shared key, and enough volume that hand-tuning a threshold breaks down. Active-learning systems earn their setup cost there. For lists where email is the key or name variants are formatting noise, a threshold does the job in seconds.
What does each cost?
Dedupe.io has historically priced by rows processed (our notes say around $9 per 1,000 rows as of mid-2026; check their site). Dedupe.dev's browser tools are free, and API fuzzy matching works out to roughly $0.19 per 1,000 rows on Starter.
Last updated