GitHub Copilot CLI gets automatic model selection: 10% discount on multipliers for all paid users

GitHub announced on April 17, 2026 in its Changelog that automatic model selection (auto model selection) for the GitHub Copilot CLI tool has become generally available (GA). The feature is now included in all Copilot plans, with the goal of reducing friction when working in the command line so that users no longer need to manually choose between Sonnet, Haiku, or GPT models for each request.

What’s changing?

Previously, Copilot CLI users had to explicitly select which model would execute their request. The new feature introduces dynamic routing — the system analyzes the request and automatically selects the most suitable model from the available pool. GitHub lists GPT-5.4, GPT-5.3-Codex, Sonnet 4.6, and Haiku 4.5 as examples in the announcement. The specific set of models depends on the user’s subscription plan and the policies set by the organization administrator.

Beyond ergonomics, auto mode also helps mitigate rate limits — if one of the frontier models is congested, the request is routed to another instead of the user receiving an error. Transparency is maintained: Copilot CLI visibly shows which model executed each individual request, and the user can exit auto mode at any time and explicitly request a specific model.

Why does this matter?

The most interesting part of the announcement concerns the billing mechanism. GitHub gives all paid users a 10% discount on the model multiplier when using auto mode. The formula is simple: a model that normally costs 1x premium request will be charged at 0.9 premium requests in auto mode. GitHub explicitly states that the discount currently applies only to models with multipliers between 0x and 1x — which covers most Copilot models, but not the most expensive frontier variants with higher multipliers.

The combination of automatic routing and 10% discount is a signal that GitHub is actively rewarding users who let the system decide which model to use. This is a sound strategy — auto routing allows GitHub better load balancing across different models and better utilization of the inference capacity of individual partners (Anthropic, OpenAI).

What’s next?

The feature is already available to all Copilot CLI users and requires no additional action — users simply need to run the latest version of the CLI tool. Organization administrators can define their own policies about which models are included in the auto pool. GitHub has already introduced a similar feature for coding agents (Claude and Codex) on the github.com platform during April 2026; extending it to the CLI rounds out coverage across all Copilot interfaces.