Copilot Usage Metrics Now Resolve Auto Model Selection to Actual Models
GitHub has updated Copilot's usage metrics dashboard to resolve the previously opaque "Auto" label into the actual model that handled each request. For organizations using auto model selection, activity that once aggregated under a generic category now breaks out by specific model — including GPT-5.3-Codex — giving admins in Copilot Business and Enterprise a granular view of which models are being used, at what volume, and by which teams across the organization.
This is a governance and cost-control win that enterprise Copilot buyers have been waiting for. As organizations move from piloting Copilot to managing it at scale, the ability to see model-level usage becomes a budget line item rather than a curiosity — especially as higher-capability models like GPT-5.3-Codex carry different cost profiles. The update also matters directly for the LTS story GitHub announced this week: enterprises that have pinned to GPT-5.3-Codex for stability and compliance reasons now have the data infrastructure to actually prove compliance with those internal policies. When AI tooling decisions start showing up in audit trails, visibility like this stops being a nice-to-have and starts being a requirement.