Used by the OpenRouter web_search server tool. Higher values return more context per search and use more search credits/tokens.
Maximum number of searches the model may perform in one request. Increase this to let the model keep searching when it needs more context.
Maximum total results across all searches in one request. Useful as a cost/context cap for agentic search.
Templates support @input and @1...@9 references. @input is replaced by your message; numbered references are replaced by the extra fields shown next to the mode selector.