GitHub Rolls Out Grok 4.6 Across Copilot Clients
The new reasoning-model option reaches five paid plan tiers and eight coding surfaces, with gradual availability, an administrator gate and usage-based billing.
GitHub began rolling out xAI's Grok 4.6 in GitHub Copilot on August 14, 2026. The model is heading to Copilot Pro, Pro+, Max, Business and Enterprise, but availability is gradual and Business or Enterprise administrators must switch on a policy that is disabled by default.
The change affects developers choosing a model for longer coding or tool-using work, as well as administrators responsible for model access and AI-credit spend. GitHub describes Grok 4.6 as a reasoning model designed for agentic coding and complex multi-step workflows, but its published performance evidence is internal rather than an independent Copilot evaluation.
Eight Copilot surfaces are in the rollout
GitHub lists Visual Studio Code, Visual Studio, Copilot CLI, Copilot cloud agent, the GitHub Copilot app, JetBrains IDEs, Xcode and Eclipse as supported destinations for the rollout. That makes this broader than a single-editor integration: the same model choice is intended to span local IDE work, command-line sessions, hosted agents and GitHub's standalone app.
The announcement does not say that every eligible account can select Grok 4.6 immediately. GitHub explicitly calls the rollout gradual and tells users to check again if the model is not visible. Its current supported-model documentation lists Grok 4.6 as generally available, while the minimum-version table still shows the recent model's IDE requirements as undetermined across Visual Studio Code, Visual Studio, JetBrains, Xcode and Eclipse. Keeping clients current is therefore the safer operational assumption during rollout.
Managed organizations get a separate access gate
Copilot Business and Enterprise administrators must enable the Grok 4.6 policy in Copilot settings before members can use it. Because that policy starts off, the model can be generally available in GitHub's catalogue yet remain unavailable inside a managed organization.
For teams, the practical checkpoint is not only whether Grok 4.6 appears in a public model list. Administrators also need to confirm that their policy permits it, and developers need to see it in the client they actually use. GitHub has not supplied adoption figures or a client-by-client completion timetable.
Usage is charged by tokens and context tier
GitHub says Grok 4.6 uses provider list pricing under Copilot's usage-based billing. Its pricing reference lists a default tier through 200,000 input tokens at $2 per million input tokens, $0.50 per million cached-input tokens and $6 per million output tokens. Above that threshold, the listed long-context rates rise to $4, $1 and $12 respectively.
Copilot converts those token charges into AI credits, with one credit equal to one US cent. Paid plans include allowances, while usage beyond the included amount is charged through additional credits where purchasing is available. The cost of selecting the model therefore depends on token volume and whether a request enters the long-context tier, not merely on the presence of a Copilot seat.
The model context is vendor-reported
xAI released Grok 4.6 on August 12 and says it focused the update on long-running agents plus more ambitious interactive and visual work. Its launch page describes a longer supplemental training run, agentic reinforcement-learning tasks and availability through its API and several partner platforms.
Those details explain why GitHub is positioning the model for sustained coding workflows, but they do not establish comparative quality inside Copilot. xAI's published benchmark table uses developer-reported or publicly available competitor results, while GitHub cites its own internal testing. No independent evidence fetched for this release measures reliability, security, tool-use accuracy or cost-effectiveness in the eight announced Copilot clients.
Status
Confirmed. GitHub's changelog and documentation establish the rollout, eligible paid plans, client list, policy gate and billing structure. Internal confidence is medium because availability is still gradual and the performance case comes from GitHub and xAI rather than independent testing.
Sources
- GitHub changelog — Grok 4.6 in Copilot
- GitHub Docs — Models and pricing for GitHub Copilot
- GitHub Docs — Supported AI models in GitHub Copilot
- xAI — Introducing Grok 4.6
Update note: Last reviewed 2026-08-14. We will revise this post if GitHub completes the rollout, publishes final client-version requirements or adds independent evaluation evidence.
Sources
- GitHub changelog — Grok 4.6 in Copilot — official
- GitHub Docs — Models and pricing for GitHub Copilot — official
- GitHub Docs — Supported AI models in GitHub Copilot — official
- xAI — Introducing Grok 4.6 — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.