Big data. Instant answers.
Pick your playground. Put search to the test.
Demo guide What to try & how it works
Full-text search over 36 million posts
Search every Stack Overflow question and answer by title and body, using words, "exact phrases" or both. Filter by tag, year and type, and every facet count and the total number of hits stays exact.
- Totals next to facets are exact counts over the whole corpus, with no "10,000+" cap: the facets visit every match anyway. Searches without facets count up to 10,000 and show "10,000+" beyond that.
- Phrase queries check word positions, which stays fast on long documents: try PMC, where an article runs to thousands of words.
- Facets are computed over the full result set, not a sample.
- TREC-COVID and FiQA come with human relevance judgments: pick a judged query to see nDCG@10 for the live ranking.
Inside this dataset
Reading the demo
Choose a suggested search or enter your own. Filters in the same group match any selected value; different groups narrow the results together. Counts reflect the current results.
Engine is query execution time. Network is the remaining wait, including the website, gateway and transfer. For vector and hybrid searches, it also includes embedding and comparison work.
Loading the live datasets...
Show the query that was sent
Elasticsearch-compatible API. With the WarpSearch API, send the same body as {"language": "es", "query": <body>} to POST /v1/collections/{collection}/search.