See for yourself.
This page shows results on different public benchmarks. We're the best on every one we've run. Toggle between them and check the evidence yourself.
Sixteen expert questions, three attempts each, on ten public FDA documents (2,198 pages) covering Saxenda, Wegovy and Zepbound. Every system answered through the same model, GPT-5.4 mini, graded by an independent judge on a majority vote across three votes. Runs 16–17 September 2026.
- 93/96 correct Two full runs of the question set (47 of 48 and 46 of 48), including the control question declined in all six repetitions.
- 100% retrieval accuracy A correct source document reached for every answerable question, every repetition.
- 40/48 most correct of five retrieval approaches, same documents, same questions, same model GPT-5.4 mini answering for every system: Hippocampus 40, OpenAI File Search 36 (average of two runs), PageIndex 33, OpenAI vector store search 29, the model alone 15.
Result deep dive
Lookups are easy for everyone; the differences are elsewhere.
Every system with document access answered most of them (19 to 24 of 24), and the model alone answered half from its own knowledge, since these are public documents. Single-fact lookups separate the systems only a little.
See it on your own documents.
Free tier, no card, no API key. Import one document and ask it the question the alternatives get wrong.