GitHub is fundamentally restructuring how enterprise teams access AI models in Copilot, moving from a one-size-fits-all approach to granular, team-level governance. The shift reflects a broader consolidation happening across the platform, with two older Google Gemini models deprecated and new administrative controls rolling out to enterprise customers starting August 3rd.

What Changed in GitHub's Copilot Model Strategy?

On July 31st, GitHub deprecated Gemini 2.5 Pro and Gemini 3 Flash across every Copilot surface, including chat, inline edits, agent modes, and code completions. The deprecation was immediate and server-side, meaning no action was required to remove the old models from most users' systems. However, the change signals something larger: GitHub is moving away from offering developers a marketplace of model choices toward a curated, policy-managed set of approved options.

The replacements are Gemini 3.1 Pro and Gemini 3.6 Flash, with Google's newest Gemini 3.6 Flash offering a notable efficiency gain. The model produces 17% fewer output tokens than its predecessor while scoring higher on developer benchmarks, meaning faster completions and lower costs for teams on usage-based billing plans.

This consolidation is part of a larger pattern. GitHub Models, a neutral API surface that let developers call any provider's model with a GitHub token, was fully retired on July 30th. Combined with the new enterprise model policy controls, the throughline is clear: Copilot is moving from a marketplace of many models to a curated, policy-managed set.

How Do Enterprise Teams Model Policies Work?

GitHub's new enterprise teams model policy targeting feature, now in public preview, gives AI administrators fine-grained control over which models their teams can access. Rather than managing model access at the organization level, admins can now set a baseline of models for the entire enterprise and then grant additional models to specific teams based on job role, function, or early experimentation needs.

The system uses a least-restrictive evaluation strategy, meaning if a user gets access to a model through any one enterprise team, they have access to that model everywhere within the company. This prevents fragmentation while still allowing flexibility.

Steps to Migrate Your Team to New Copilot Models

• Check Your Current Model Selection: Open the model selector in VS Code and Copilot Chat to confirm which Gemini models are actually listed. If you pinned Gemini 2.5 Pro or Gemini 3 Flash, switch to Gemini 3.1 Pro or Gemini 3.6 Flash respectively.
• Enable Preview Models in Enterprise Settings: If you are on Copilot Enterprise and Gemini 3.1 Pro is not visible, have an admin check the model policy in Copilot settings and enable the preview model. Gemini 3.1 Pro is a preview model and is commonly disabled by default in enterprise policies.
• Update Configuration Files and Documentation: Search your repository for the old model strings, including any in configuration files or team documentation, and update anything that instructs the model to use the deprecated names.
• Test Before Full Rollout: Run a quick smoke test on your most important assisted workflow with the new model before relying on it. If you automated prompts that depended on Gemini 2.5 Pro's specific behavior, budget a small evaluation pass to ensure compatibility.

Why Is GitHub Making These Changes Now?

The timing reflects Microsoft's broader AI monetization strategy. Azure surpassed $100 billion in annual revenue while quarterly growth accelerated to 43%, with commercial remaining performance obligations surging 84% to $678 billion. This revenue visibility is supported by broad enterprise AI adoption, and GitHub Copilot is a key part of that ecosystem.

By consolidating the model landscape and introducing team-level governance, GitHub is making Copilot more enterprise-friendly. Companies can now tailor AI tool access by role without deferring entirely to organization-level decisions. This addresses a real pain point: historically, model access has been managed at the organization level, but enterprise teams give admins finer-grained, user-based control that better maps to how people actually work.

The shift also reduces complexity. For individual developers, the practical cost is low, mostly a model-selector change. For teams that built scripts, prompts, or automation around specific model strings, this is a reminder that model names in Copilot are now moving targets. However, the newer models are generally more efficient and capable, so the migration is typically a net positive.

What Models Survive the Deprecation?

Google's current lineup in Copilot includes three models: Gemini 3.6 Flash and Gemini 3.5 Flash at general availability status, plus Gemini 3.1 Pro in public preview. Notably, Gemini 3.5 Flash-Lite, launched on the same day as Gemini 3.6 Flash, already outscored the deprecated Gemini 3 Flash on multiple benchmarks despite costing significantly less. This suggests the deprecation is not a downgrade for most workloads.

The rollout of enterprise teams model policy targeting is happening gradually, with most enterprise customers gaining access to the preview opt-in on August 3rd. During the preview period, companies can roll back to reset their policy to its previous configuration if needed, giving teams time to evaluate the new governance model before fully committing.