Short answer: In an interactive Copilot CLI session, run /model to see the models available to your account and choose one. That list—not a model list from another IDE, another user, or an older article—is the practical answer to “Which models can I use?” Availability can differ by Copilot plan, client, and organization or enterprise settings, and can change over time.
See and switch models in Copilot CLI
- Start an interactive session by running
copilotin your terminal. - Enter
/model. - Use the up and down arrows to browse the models shown. Press Enter to select one. If a model offers context or reasoning options, the picker also lets you change those settings.
GitHub documents that a model choice applies for the rest of the current session and persists if you resume that session; it does not affect other sessions. Use the picker in the session where you want to change models. (GitHub Docs: Choosing a model in Copilot CLI)
There isn’t one reliable, permanent list that applies to every Copilot CLI user. GitHub says model access depends on your Copilot plan and organization or enterprise settings. Models can also differ by Copilot client and change over time. Use the picker for your account, and GitHub’s supported-models documentation for the broader, changing catalog.
A dated example—not a current catalog: On March 25, 2026, one Copilot Pro subscriber using CLI version 1.0.11 reported seeing Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Haiku 4.5, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4, Gemini 3 Pro (Preview), GPT-5.4, GPT-5.3-Codex, GPT-5.2-Codex, GPT-5.2, GPT-5.1-Codex-Max, GPT-5.1-Codex, GPT-5.1, GPT-5.4 mini, GPT-5.1-Codex-Mini (Preview), GPT-5 mini, and GPT-4.1. That is one person’s account-specific snapshot, not a guarantee that those models are available to you now. The discussion is marked unanswered.
Choose by task, then test on your own work
Start with the job you need done, rather than assuming one model is best at everything. GitHub’s model comparison offers task-oriented starting points, including GPT-5.3-Codex for agentic software development, GPT-5.6 Luna for quick help with simple or repetitive tasks, GPT-5.6 Sol for deep reasoning and debugging, and GPT-5.6 Terra for general-purpose coding and agent tasks. These are starting points, not guarantees of performance; use them only if the model appears in your CLI picker.
A practical selection process:
- Small, routine change: Try a model GitHub positions for quick or simple tasks. Check whether it follows instructions and completes the change without unnecessary steps.
- Complex debugging or multi-file work: Try a model described for deeper reasoning or agentic development. Check whether it identifies the root cause, makes coherent edits, and runs or proposes relevant tests.
- Long context or large repository: If the picker offers context tiers, compare the default and extended option for your task. A larger context is a capacity choice, not a guarantee of a better answer.
- Not sure which model to choose: If Copilot Fusion appears in your CLI picker, you can select it instead of a specific model. GitHub describes Fusion as a system that chooses models and execution patterns for a task; it is not a single fixed model. (GitHub Docs: Choosing a model in Copilot CLI; About GitHub Copilot Fusion)
For a reproducible comparison, give two available models the same representative task in a clean branch or disposable worktree. Compare correctness, instruction following, test results, tool use, latency, and any usage cost that applies to your plan. Review the diff before keeping changes. This is a way to evaluate models in your own environment—not a claim that one named model will win.
GitHub’s billing terms also matter when comparing options. Its current Copilot pricing documentation describes usage in GitHub AI Credits, with rates based on model and tokens; 1 AI credit equals $0.01 USD, and included allowances vary by plan. Those are Copilot billing terms, not a quote for using a model provider’s API directly. Don’t assume an older claim such as “every prompt costs the same” applies across models or billing arrangements. (GitHub Docs: Models and pricing for GitHub Copilot)
Why a model may show in an IDE but not in CLI
An IDE’s model picker does not prove that the same model is enabled in Copilot CLI. GitHub notes that availability depends on the Copilot surface as well as plan and organization or enterprise settings. A model missing from CLI does not, by itself, show that your subscription is inactive or misconfigured. (GitHub Docs: Models in GitHub Copilot)
A GitHub Community user reported seeing Grok, Gemini 3 Flash (preview), and GPT-4o in PyCharm but not in their CLI list; entering /model GPT-4o reportedly returned Model GPT-4o is unsupported. A commenter suggested differences between clients, rollout stage, or backend configuration, and suggested checking the CLI version and subscription. Those were community hypotheses and suggestions, not an official diagnosis; the discussion did not report a confirmed resolution.
If a model you expect is missing, try these checks:
- Inspect the CLI picker: Run
/modeland use the models actually listed there. The supported-models page is broader than an account-specific guarantee. - Check your CLI version: If you suspect the CLI is outdated, check its version and follow the update instructions for your installation method. A community commenter recommended using the latest version but did not specify an update command or report a resolution.
- Check organization settings: If you use Copilot through an organization or enterprise, ask your administrator whether the model is enabled for your account. GitHub says administrator settings can affect model access.
- Compare the correct plan and client: Check availability for your plan and Copilot CLI specifically—not solely against another IDE.
If it still looks wrong, note your CLI version, plan, organization status, the exact model name, and the error message before reporting the issue. Don’t treat an IDE listing or an older user’s model list as proof that your CLI must support that model.
Copilot CLI model selection is not production API access
Selecting a GitHub-hosted model in Copilot CLI does not establish that you have a production API entitlement, a provider API key, or permission to call that model from your own application. The CLI’s /model command selects a model for a Copilot session; it is not a general model-inference API for an application.
There is a related but distinct option: GitHub documents bring your own key (BYOK) for connecting Copilot CLI to a compatible external provider or model endpoint. That setup uses credentials you supply and the provider endpoint you configure. GitHub lists OpenAI-compatible endpoints, Azure OpenAI, and Anthropic among the supported provider types; models must support streaming and tool calling for CLI use. This is a way to use an external model through the CLI, not evidence that a Copilot subscription grants production API access or pays the provider’s API bill. (GitHub Docs: Adding LLM models to GitHub Copilot CLI)
For a production application, verify the intended model directly in the provider’s official API and commercial documentation. Confirm the endpoint, authentication, model identifier, version or deprecation policy, region restrictions, rate limits, and billing unit and currency. Keep that API’s access and cost separate from Copilot’s model picker and GitHub AI Credits. The sources cited here do not establish production API availability or pricing for every model shown in Copilot CLI.
Quick decision guide
| If you need… | Start here | Verify before relying on it |
|---|---|---|
| A model for a CLI task | Run /model and select an available option |
The picker is account- and client-specific |
| A task-based starting point | Use GitHub’s model comparison for general coding, quick tasks, or deep debugging | The suggested model must also appear in your CLI picker |
| A routed choice instead of one fixed model | Select Copilot Fusion if it appears in your CLI picker | Fusion chooses models and execution patterns; it is not one fixed model |
| A missing model explained | Check the CLI version and, for work accounts, ask about administrator model settings | IDE availability alone does not confirm CLI access |
| A model in your production app | Check the provider’s official API documentation and obtain the appropriate access | CLI access, Copilot billing, and provider API billing are separate |
Sources
- GitHub Docs: Choosing a model in GitHub Copilot CLI
- GitHub Docs: Supported AI models in GitHub Copilot
- GitHub Docs: AI model comparison
- GitHub Docs: Models in GitHub Copilot
- GitHub Docs: About GitHub Copilot Fusion
- GitHub Docs: Adding LLM models to GitHub Copilot CLI
- GitHub Docs: Models and pricing for GitHub Copilot
- GitHub Community: Some models not available in Copilot CLI with Pro (user report; unanswered)
Sources
Choosing a model in GitHub Copilot CLI
Supported AI models in GitHub Copilot
Adding LLM models to GitHub Copilot CLI