GitHub Copilot Usage Metrics: Auto Model Selection Resolved to Actual Models
GitHub CopilotGitHub Copilot's usage metrics now surface the actual AI model name behind every auto mode interaction, replacing the previously unhelpful generic "Auto" label. This change applies across the REST API and the dashboard — including totals_by_model_feature breakdowns at the enterprise, organization, and user levels. Organizations relying on auto model selection as their default can now gain precise, auditable insight into which models their teams are actually consuming. The update is live now for API requests using version 2026-03-10 or later.
Sources & Mentions
4 external resources covering this update
Overview
GitHub has updated Copilot usage metrics to resolve auto model selection activity to the specific model that was actually used, closing a significant visibility gap for organizations that have adopted auto mode as their default configuration.
What Changed
Previously, when Copilot's auto model selection feature chose a model on behalf of a user, the resulting activity appeared in usage metrics under a generic label rather than the model's actual name. This made it difficult for administrators to understand their organization's true model consumption patterns.
With this update, all auto mode activity is attributed to the underlying model that served the request. The change flows through the entire metrics surface:
REST API
The totals_by_model_feature field in the enterprise, organization, and user-level report endpoints now shows the resolved model name. Organizations querying the API on version 2026-03-10 or later will see the updated attribution automatically.
Dashboard
The "Model usage per chat mode" charts in the Copilot usage metrics dashboard now display per-model breakdowns that include auto-mode activity, attributed to the actual model used.
Impact for Administrators
For organizations that have enabled auto model selection broadly — particularly on Copilot Business and Copilot Enterprise — this change significantly improves the quality of usage data. Administrators can now:
- Identify which models are being selected most frequently by the auto system
- Perform accurate compliance and audit reporting at the model level
- Understand cost and consumption patterns more precisely as auto mode adoption scales
Current Limitation
Auto usage counts roll into the resolved model's totals. A separate breakdown comparing auto-mode requests versus manually selected model requests is not yet available. GitHub has indicated this type of split reporting may be introduced in a future update.
Availability
The improvement is live across all Copilot plans. The REST API update requires API version 2026-03-10 or later.