Grok 4.6 became selectable in Copilot on 14 August. That is the fourth model added in nine days, after Kimi K3 on 6 August, MAI-Code-1.1-Flash on 11 August and Gemini 3.7 Flash on 13 August.
What separates this one is the pitch. GitHub is not selling it on general quality — it is positioning it for agentic coding and complex multi-step workflows, citing internal testing on lengthy tasks that require sustained reasoning.
Where you can select it
Available on Copilot Pro, Pro+, Max, Business and Enterprise, across most surfaces:
- Visual Studio Code, Visual Studio, JetBrains, Xcode and Eclipse
- The Copilot CLI and the Copilot cloud agent
- The GitHub Copilot app
Off by default on Business and Enterprise
GitHub is explicit: Copilot Enterprise and Copilot Business plan administrators must enable the Grok 4.6 policy in Copilot settings, and the policy is off by default.
This is now the fourth time in nine days that sentence has appeared in a changelog. If you administer a plan, the practical consequence is that model enablement has become a recurring administrative task rather than a one-off — and the queue of unanswered ones is growing.
The rollout is also gradual, so availability varies by user. An absent entry on day one is not a fault to report.
What "agentic" means for your bill
Grok 4.6 is billed at provider list pricing under usage-based billing. GitHub does not state a premium request multiplier, so treat that as unstated rather than assuming parity with anything else.
The positioning and the billing model interact in a way worth thinking about before you make this a default. A model marketed for long, multi-step agent runs is a model designed to consume a lot of tokens per task, and usage-based billing charges by exactly that. A short chat with an expensive model is cheap; an hour-long agent run with a cheap one is not.
The per-model token breakdown GitHub shipped on 11 August is the instrument for checking this. If you enable Grok 4.6, look at its output and input columns after a fortnight rather than at the end of the quarter.
What GitHub does not claim
The entry does not say what changed relative to Grok 4.5, which reached Copilot on 28 July. It highlights strengths in complex workflows and tool use without offering a comparison to the version it succeeds.
That absence is itself useful information. When a vendor ships a point release and does not describe the delta, the honest reading is that the difference is not large enough to summarise in a sentence — which means your own three-task evaluation will tell you more than the announcement does.