GitHub Copilot: Auto Model Selection Now Generally Available in JetBrains IDEs

GitHub Copilot Auto Model Selection Reaches General Availability in JetBrains IDEs

GitHub has graduated auto model selection for GitHub Copilot in JetBrains IDEs from public preview to general availability, extending the feature to all Copilot subscription tiers as of March 12, 2026. The feature had been in preview since November 2025, and its GA promotion marks a significant milestone for JetBrains users who previously had to manually switch between AI models.

How Auto Model Selection Works

Rather than requiring developers to choose a specific model for each task, the auto selection mode allows GitHub Copilot to make that decision dynamically. The system evaluates real-time model availability and service performance to route each request to the most appropriate option from a pool that currently includes GPT-5.4, GPT-5.3-Codex, Claude Sonnet 4.6, and Haiku 4.5. The specific models in the pool may evolve over time as GitHub adds or removes models.

The dynamic routing approach is designed to mitigate rate limits — if one model is under heavy load, Copilot can route the request to an alternative without any manual intervention from the developer. Administrators retain full control: auto selection honors all organizational model policies, ensuring that model restrictions set by IT or engineering leadership are respected even when auto is enabled.

Transparency and Developer Control

GitHub has built explicit transparency into the feature. Developers can see which model processed any given response by hovering over it in the IDE, removing uncertainty about which AI produced a particular suggestion. Manual model selection remains available at any time — auto is the default when selected, but switching to a specific model requires only a single interaction.

Billing: A 10% Discount for Paid Subscribers

For teams and individual developers on paid Copilot plans, auto model selection carries a concrete financial benefit. GitHub applies a 10% discount on model multipliers when auto selects a model, rather than the user explicitly choosing it. For example, a model normally priced at a 1.0x premium request multiplier costs 0.9 premium requests when selected by auto. This incentive is designed to encourage adoption of the auto mode.

Future Direction

GitHub has signaled that the auto selection capability will become more sophisticated over time. Future versions are expected to match model selection to task complexity — routing lightweight requests to faster, smaller models and directing complex reasoning tasks to higher-capacity models. This would shift auto from a primarily availability-based routing mechanism to a more intelligent, task-aware system.