GitHub deprecated MAI-Code-1-Flash across GitHub Copilot experiences on September 10, 2026. The change affects more than a chat model picker: GitHub identifies Copilot Chat, inline edits, ask mode, agent mode, and code completions among the affected experiences.
GitHub recommends MAI-Code-1.1-Flash as the replacement. Although the company says users do not need to take action to remove the deprecated model, developers and administrators should still review explicit model selections in repositories, automation, Copilot CLI configurations, and enterprise policies. The available documentation does not establish that those references will be migrated automatically.
What changed in GitHub Copilot
The deprecation applies across GitHub Copilot experiences rather than to one client or interaction mode. Existing workflows that refer to MAI-Code-1-Flash may therefore include interactive development features as well as scripts and integrations.
GitHub’s deprecation announcement points users toward MAI-Code-1.1-Flash and says the replacement will appear in the Copilot Chat model selector in Visual Studio Code and on GitHub.com after it is enabled. The announcement does not document an automatic rewrite of existing model identifiers.
The supported-models reference lists MAI-Code-1.1-Flash as generally available on GitHub.com, Copilot CLI, Visual Studio Code, Visual Studio, Eclipse, Xcode, and JetBrains IDEs. It also lists the model for Copilot Pro, Pro+, Max, Business, and Enterprise plans. Availability can still vary according to the user’s plan, client, usage context, and administrative settings.
Availability depends on client and policy settings
General availability does not guarantee that every Copilot user will see or be able to select MAI-Code-1.1-Flash. The supported-models documentation lists Visual Studio Code 1.121 as the minimum listed VS Code version for the replacement. It lists no minimum version for Visual Studio, JetBrains IDEs, Xcode, or Eclipse for this model.
Enterprise administrators have additional controls. According to GitHub’s enterprise model-availability documentation, enterprise owners can set model availability for everyone, delegate decisions to organizations or enterprise teams, create targeted organization rules, and configure model access separately for agent apps.
Those controls can enable or disable generally available models by default. An enterprise-level restriction can also prevent a team from enabling a model that the enterprise has disabled. Administrators should therefore check enterprise, organization, team, and agent-app policies before assuming that the replacement is available to all affected users.
Review explicit Copilot CLI model selections
Copilot CLI is an important migration surface for engineering teams that use scripts, CI/CD pipelines, or other automation. The Copilot CLI reference documents several ways to select a model:
- The
--modelcommand-line option - The
COPILOT_MODELenvironment variable - The
modelkey in~/.copilot/settings.jsonor$COPILOT_HOME/settings.json - Custom-agent definitions
The documented precedence is custom agent, command-line option, environment variable, persistent configuration file, and then the CLI default. This ordering matters when an old model identifier appears in more than one place: changing a lower-precedence setting may not affect the model actually selected by a custom agent or command invocation.
The documentation identifies these configuration surfaces but does not specify what happens when a request continues to use MAI-Code-1-Flash after retirement. In particular, the supplied documentation does not establish whether such requests will fail, fall back to another model, or receive a grace period.
Model changes are experience-specific
Developers should not assume that selecting MAI-Code-1.1-Flash in one Copilot interface changes every other Copilot feature. GitHub’s inline-suggestions documentation states that changing the model for inline suggestions affects Copilot ghost-text suggestions. It does not affect next edit suggestions or Copilot Chat.
This distinction makes validation important. A replacement selected for inline suggestions is not evidence that Chat, ask mode, agent mode, or automation is using the same model. The available model list can also change over time, and Business and Enterprise users may need administrator enablement before the model switcher is available.
Pricing information for the replacement
GitHub’s models and pricing reference lists MAI-Code-1.1-Flash as a generally available lightweight model priced at $0.20 per 1 million input tokens, $0.02 per 1 million cached input tokens, and $1.20 per 1 million output tokens.
That documentation also says code completions and next edit suggestions are not billed in AI credits. They remain unlimited for paid Copilot plans and retain their existing counting mechanism. The supplied documentation does not provide comparable MAI-Code-1-Flash pricing, so the replacement’s listed rates should not be treated as proof of a before-and-after price change for a particular workflow.
What developers and administrators should do
- Search for explicit model references. Review source repositories, CI/CD definitions, scripts, custom agents, CLI invocations, and persistent settings for
MAI-Code-1-Flash. - Update supported references. Replace explicit references with
MAI-Code-1.1-Flashwhere the replacement is available and permitted by the relevant plan and policy. - Inspect Copilot CLI precedence. Check
--modelarguments,COPILOT_MODEL, custom-agent definitions, and~/.copilot/settings.jsonor$COPILOT_HOME/settings.json. - Review enterprise controls. Confirm that enterprise, organization, team, and agent-app policies allow the replacement for the users and workflows that need it.
- Verify client availability. Check the model selector or relevant Copilot settings, including the documented VS Code 1.121 minimum listed for MAI-Code-1.1-Flash.
- Test affected workflows. Validate Chat, inline edits, ask mode, agent mode, code completions, and automation after changing model references. Watch for unsupported-model errors or behavioral changes because the retirement-time response is not documented.
What remains undocumented
The evidence confirms the deprecation date and recommended replacement, but it does not document automatic migration of existing selections. It also does not define the exact behavior of requests that continue to identify MAI-Code-1-Flash.
There is no supplied comparison of the two models’ context limits, rate limits, performance, quality, output compatibility, or capabilities. Developers should avoid assuming that prompts, outputs, latency, or limits will remain unchanged after switching. The change is not described as a security incident or vulnerability, and the supplied sources do not identify new security or data-collection behavior caused by the replacement.
