search-benchmark

You can benchmark cross-model search quality by dispatching identical scenarios to opus, sonnet, and haiku in parallel. This measures search suggestions equally across model tiers and produces a structured comparison report.

Cross-model search benchmarking is used to evaluate search tool quality for tool discovery. It identifies performance differences across model tiers and assesses Phase 2 tag impact to reveal optimization opportunities.

Yes, you can compare search results across opus, sonnet, and haiku simultaneously. The benchmark dispatches identical scenarios to all three models in parallel to ensure equal treatment and generate a comparison table.

You generate a comparison report for multimodel search results by running a benchmark that applies synthetic and live-model scenarios across opus, sonnet, and haiku. It collects the results and outputs a structured comparison table.

Yes, the search benchmark supports both synthetic and live-model scenarios. It applies these scenarios equally across opus, sonnet, and haiku to compare results and measure cross-model search quality.

A limitation of multimodel search benchmarking is that it requires equal treatment across opus, sonnet, and haiku to compare results accurately. It focuses strictly on search tool quality and Phase 2 tag impact rather than general model performance.