GitHub Copilotにおける2026年8月のモデル廃止予定Upcoming August 2026 model deprecations in GitHub Copilot

GitHub has announced that several artificial intelligence models will be retired across the full range of GitHub Copilot experiences on September 1, 2026, a change that will require affected developers and organizations to switch to supported alternatives before that date. The move matters because Copilot is embedded in the daily coding workflows of a large number of developers, and a model deprecation can alter the behavior, availability, and configuration of features that teams rely on.

According to the changelog notice, the deprecation spans every major surface where Copilot operates. That includes Copilot Chat, inline edits, the ask and agent modes, and code completions. In practical terms, the change is not limited to a single feature such as conversational assistance; it reaches the autocomplete suggestions that appear as developers type, as well as the more autonomous agent workflows that can plan and execute multi-step tasks.

Model deprecations are a routine part of maintaining an AI-powered platform. GitHub Copilot draws on models from multiple providers, and the underlying model catalog changes regularly as vendors release newer versions and retire older ones. When a provider sunsets a model or a newer generation supersedes it, GitHub typically removes the older option and points users toward a replacement. The company's changelog posts usually list the specific models being deprecated alongside their individual deprecation dates, so administrators can identify exactly which options will disappear.

For most users, the transition is likely to be straightforward. Copilot includes a model picker that lets users select which model powers Chat and related features, and when a chosen model is deprecated, requests are generally redirected to a supported default. Developers who have not manually pinned a specific model may not notice any disruption, while those who rely on a particular deprecated model will need to choose an alternative. Organizations that manage Copilot through policy controls should review their settings to confirm that the models they permit remain available after the cutoff.

The impact appears likely to be more significant for teams that have standardized on a specific model for consistency, compliance, or performance reasons. Different models can vary in response style, latency, context window size, and cost, so a forced migration may prompt some evaluation before September 1. Enterprises using Copilot in regulated environments, in particular, may want to validate that a replacement model meets their internal requirements and produces acceptable results across their codebases.

This announcement fits within a broader industry pattern. The pace of large language model releases has accelerated, and the tools built on top of them, including coding assistants, must continuously rotate their supported models to keep current. Competing products such as Amazon's coding assistants, Google's Gemini-based developer tools, and various independent editors integrated with models through APIs face the same maintenance cycle. Frequent deprecations are one of the trade-offs of building on rapidly evolving foundation models, and providers generally try to give advance notice so downstream users can plan.

GitHub's approach of announcing the change roughly a month ahead, with the notice published at the end of July 2026 for a September 1 effective date, gives teams a window to test alternatives and update any automation or documentation that references the affected models. Users who interact with Copilot only through default settings will typically be moved automatically, but it is worth confirming which specific models are named in the official changelog entry, since that list determines who is affected.

To prepare, developers and administrators can review the full list of deprecated models in GitHub's changelog, identify whether any are in active use, and select supported replacements ahead of the deadline. Checking the model picker in Copilot Chat, reviewing organizational policies for enabled models, and testing key workflows against a replacement model are practical steps. Because Copilot continues to add and remove models over time, treating model selection as a periodically reviewed setting, rather than a one-time choice, is likely to reduce friction when future deprecations occur.

The core takeaway is narrow but important: the affected models will stop working across all Copilot experiences on September 1, 2026, and anyone depending on them should identify a supported alternative before that date to avoid interruption.