{"version":"2.1.285","anchor":"headless-latency-telemetry-adds-per-query-round-overhead-met","canonical_anchor":"headless-latency-telemetry-adds-per-query-round-overhead-met","heading":"Latency telemetry for non-interactive runs now measures each round","tier":"internal","area":"Telemetry","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.285\/e\/headless-latency-telemetry-adds-per-query-round-overhead-met","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.285","markdown":"### Latency telemetry for non-interactive runs now measures each round\n\nSpeed data from non-interactive runs now records each round of model requests separately, not just one total per turn\n\n**Unclear.** It is not clear whether this data is sent for every run or only for a sample.\n\n**What**\n\nTelemetry is the usage data Claude Code sends back to Anthropic. A headless run is one without the interactive screen, for example when a script drives Claude Code. Speed data from these runs used to record one overhead figure and one time-to-first-response per turn. It now also records figures for each round of requests within a turn, including:\n\n- `query_round_count`, the number of rounds\n\n- `query_overhead_total_ms`, the total overhead in milliseconds\n\n- `time_to_first_response_first_round_ms`, how long the first round took to start answering\n\nThe event recorded after a successful model request also gains an overhead figure and a breakdown of time not covered elsewhere.\n\n**Why**\n\nThis lets Anthropic measure delays in turns that make several requests to the model. It does not change how Claude Code behaves for you.\n\n- Area: Telemetry\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 1\/5"}