GitHub is gradually rolling out Grok 4.7, which it describes as xAI’s latest reasoning model, in GitHub Copilot. The model is intended for agentic coding and complex, multistep workflows, giving eligible developers another option to evaluate alongside the models already available in Copilot.
Grok 4.7 is not arriving for every user at the same time. Availability depends on the Copilot plan, the client or development surface in use, and— for Business and Enterprise deployments—organization or enterprise model policies. Billing also depends on the account’s pricing model, particularly for existing annual Pro and Pro+ subscribers.
What changed in GitHub Copilot
In its September 21, 2026 announcement, GitHub said Grok 4.7 was beginning its rollout in Copilot. The eligible plan types named in the announcement are Copilot Pro, Pro+, Max, Business, and Enterprise.
Where the rollout has reached a supported client, users can select Grok 4.7 from the Copilot model picker. GitHub identifies the following Copilot surfaces in connection with the rollout:
- Visual Studio Code
- Visual Studio
- Copilot CLI
- GitHub Copilot cloud agent
- GitHub Copilot app
- JetBrains
- Xcode
- Eclipse
The announcement does not establish that all of these surfaces receive the model simultaneously. It also does not provide a completion date for the gradual rollout, so a user on an eligible plan may not see Grok 4.7 immediately.
Who can access the model
For individual Copilot plans, the configuration guidance says available AI models do not require separate model-access configuration. Users should instead check the model picker in a supported Copilot surface. Copilot Free and Student accounts are limited to automatic model selection according to GitHub’s configuration documentation, and they are not among the plan types identified for this Grok 4.7 rollout.
Business and Enterprise users may have an additional administrative dependency. Organization and enterprise owners can enable or disable model access through Copilot model settings, while enterprise-enforced policies can prevent organization owners from changing those settings. An organization’s default availability policy can also determine whether newly released models are enabled when they have not been configured individually.
GitHub’s supported-model catalog lists Grok 4.7 and identifies xAI as its provider. The catalog explains that model availability can vary according to the plan and client being used. The available documentation does not provide a reliable, complete Grok 4.7 client-by-client compatibility matrix or a definitive minimum version for every surface, so administrators should verify access in the clients their teams actually use.
Billing depends on the account’s pricing model
Under GitHub’s current usage-based billing documentation, Grok 4.7 costs:
- $2.00 per million input tokens
- $0.50 per million cached input tokens
- $6.00 per million output tokens
Those rates apply to the default context tier up to 200,000 tokens. For interactions above 200,000 tokens, the documented long-context rates are $4.00 per million input tokens, $1.00 per million cached input tokens, and $12.00 per million output tokens. GitHub states that one AI credit equals $0.01 USD and that usage-based costs depend on the selected model and the tokens consumed.
There is an important exception for existing annual Copilot Pro and Pro+ subscribers. Those who remained on legacy billing after June 1, 2026 can continue using premium request-based billing until their annual plan ends. That system uses model multipliers rather than the current token-based calculation. As a result, the token prices above should not be treated as the billing method for every eligible Pro or Pro+ account.
Teams planning extended agentic or long-context tasks should review both their available AI-credit allowance and the pricing model attached to their Copilot subscription before making Grok 4.7 part of a regular workflow.
Why the rollout matters to developers
GitHub positions Grok 4.7 for agentic coding and complex multistep workflows. That makes it relevant for developers evaluating models that can participate in tasks extending beyond a single completion, such as longer debugging or implementation workflows. The availability of the model across editor, IDE, command-line, cloud-agent, and app surfaces also allows teams to assess it in the environments where they already work.
However, the supplied documentation does not establish that Grok 4.7 is faster, more accurate, or better than Grok 4.6 or any other Copilot model. It also does not include Grok 4.7-specific benchmark results or latency measurements. Teams should therefore treat this release as an opportunity for workflow evaluation rather than evidence of comparative performance.
Availability and data-residency caveats
Regular Copilot availability and data-residency deployments should not be assumed to have identical model sets. GitHub says that data-residency deployments expose only models certified and available in the designated region, and that newly released models may take longer to arrive because of regional infrastructure and certification requirements.
The supplied regional documentation does not list Grok 4.7 among its examples, so it does not establish that the model is available in any particular data-residency region. Organizations using regional enforcement should confirm availability through their applicable Copilot configuration rather than assuming that the general rollout applies to them.
GitHub’s documentation describes deployment-wide routing, inference processing, logs, and telemetry restrictions for data-residency policies when those policies are enabled. That is not a Grok 4.7-specific security, privacy, retention, or compliance guarantee, and the supplied sources do not document such guarantees for this model.
What you should do
- Check the model picker. In a supported Copilot surface, look for Grok 4.7. If it is not listed, the gradual rollout may not have reached that account or client yet.
- Review organization policies. Business and Enterprise administrators should inspect organization and enterprise model settings if users cannot access the model. Check whether an enterprise-level policy overrides organization settings.
- Confirm billing treatment. Review the account’s AI-credit allowance and token pricing before running substantial workloads. Existing annual Pro and Pro+ subscribers should also determine whether they remain on legacy premium-request billing.
- Evaluate representative tasks. Compare Grok 4.7 with currently available models using the debugging, coding, and multistep workflows that matter to your team. Do not assume performance advantages that GitHub has not documented.
- Check regional constraints. If your organization uses data-residency enforcement, verify that Grok 4.7 is certified and available in the designated region.
For now, the practical change is an additional model choice for eligible Copilot users, not a universal switch in how Copilot selects or bills models. Access, cost, and regional availability still need to be checked in the context of each account and deployment.
