Elicit tops biomedical search benchmark with 60.3% recall, 20 points ahead of rivals

elicitorg · x · 2026-09-05

Elicit published a search API benchmark on BioASQ (5,486 biomedical questions) comparing itself to Exa, Perplexity, Brave, and Parallel. At 50 results, Elicit recalled 60.3% of gold-standard papers vs 40.3% for the next best web search API — roughly 50% more relevant literature. Note: this is Elicit's own self-run evaluation; methodology linked in the post.

Related event: Elicit Tops BioASQ with 60.3% Recall, Beating Web Search by 20 Points(3 posts)→

Original post →

More from Apps

Apps channel →