Skip to main content
We’re excited to introduce dynamic snippets in the Search endpoint. Our models read the whole page at query time and keep only the parts relevant to your query.

What changes

  • Smaller snippets: each result’s snippet is about half the size, so your agent reads fewer tokens on every call.
  • Better accuracy: the passages that answer your query stay in, and results are more accurate.
  • Same response format: snippets are still returned in the content field of each result.

Rollout

Dynamic snippets are rolling out by default at every depth during the week of November 16, 2026. There is nothing to change in your requests, and the price stays the same. For more details, see Search overview.