Originally published at https://tekmag.thsite.top/github-copilot-now-ships-xai-grok-models-what-multi-vendor-model-choice-means-for-developers/
GitHub Copilot is an AI coding assistant that now offers a model picker spanning six vendors, and xAI's Grok 4.5 and Grok 4.6 have arrived in that picker as generally available models, alongside OpenAI's GPT-5 series, Anthropic's Claude family, Google's Gemini line, Moonshot's Kimi models, and Microsoft's MAI-Code builds.
Key Takeaways
- Grok is generally available: GitHub's supported-models documentation lists Grok 4.5 and Grok 4.6 as GA entries in the Copilot model selector.
- Six vendors, one picker: the current roster spans OpenAI, Anthropic, Google, Moonshot, Microsoft, and xAI, from GPT-5 mini to Claude Opus to Gemini 3.8 Flash.
- The older Grok entry is gone: Grok Code Fast 1 was deprecated on May 15, 2026, and GitHub's changelog points users to the newer versions or to GPT-5 mini.
- Version gates apply: Grok 4.5 requires Visual Studio 17.14.19 at minimum, and the floor for Grok 4.6 is still TBD in the documentation.
- The rollout is gradual: there is no single "Grok added" entry in the September 2026 changelog. Model additions have been rolling in without one launch day.
What Multi-Vendor Model Choice Means for Developers
Multi-vendor model choice means picking a model per task instead of a vendor per tool. One IDE session can run GPT-5 mini for fast completions, Claude Opus for long-horizon reasoning, and Grok 4.5 for specialized coding, without leaving the editor. The model is now a per-task setting, in the same role that shell profiles and build targets used to play.
According to GitHub's supported-models reference for Copilot, the current roster covers six vendors. OpenAI contributes the GPT-5 series, including mini, Codex, and the 5.6 line with its Luna, Sol, and Terra builds. Anthropic contributes the Claude family, listed as Fable, Haiku, Opus, and Sonnet 5. Google contributes the Gemini line, including 3.8 Flash. Moonshot contributes Kimi K2.7 Code and K3. Microsoft contributes MAI-Code. And xAI contributes Grok 4.5 and 4.6. That is more than a dozen model entries in a single selector, and the count grows with every changelog cycle.
The Grok Lineup in Copilot, Version by Version
Copilot now carries two xAI models, and both are generally available. GitHub's Copilot changelog records the full history: the earlier entry, Grok Code Fast 1, was deprecated on May 15, 2026, with developers told to move to the newer versions or to GPT-5 mini. Grok 4.5 and Grok 4.6 replaced it as the supported xAI entries, and both sit in the model selector with GA status as of September 2026.
The documentation attaches real constraints to each model. Grok 4.5 requires Visual Studio 17.14.19 at minimum, so any install older than that build will not show the model in the picker at all. The minimum IDE version for Grok 4.6 is listed as TBD, which means the model is live before its documentation entry is complete. For the release history behind those version numbers, the Visual Studio release notes are the reference to check against.
How a Single-Vendor Tool Became a Six-Vendor Platform
Copilot's early years were OpenAI-only, and multi-vendor choice was not part of the launch. By September 2026 the inverse is true: according to GitHub's supported-models reference, five other vendors sit beside OpenAI in the same selector, and adding a model is now a routine maintenance step rather than a partnership announcement. The shift happened through changelog entries, not launch events.
GitHub's changelog backs that up. Recent cycles list Gemini 3.8 Flash and GPT-6 Astra at GA status, and there is no "Grok added" entry anywhere in September 2026, which is why social posts describing a Grok news day for Copilot do not hold up. The pattern is sustained rollout. xAI's Grok model series, built by the lab that develops it, has been entering the Copilot lineup through that same gradual process, one model family at a time.
Model Per Task: Matching a Model to the Work
The working pattern is to match the model to the task. Fast autocomplete wants low latency, long-horizon reasoning wants a heavyweight model, and specialized coding logic wants a model with strong code benchmarks. Multi-vendor choice turns that matching into a setting inside the IDE instead of a reason to buy a second subscription or switch tools.
The task mapping that falls out of the current roster: Gemini 3.8 Flash and GPT-5 mini for speed and latency, Claude Opus and the GPT-5.6 series for complex reasoning, and Grok 4.5, 4.6, and Kimi K3 for specialized coding and logic. The fourth benefit is versatility, which is the ability to switch models mid-workflow, compare two outputs on the same prompt, or route around a limitation in one model by jumping to another. The model race extends past the IDE itself. A recent roundup of the top GitHub repos, including Claude plugins and free Codex alternatives, shows the surrounding tool layer moving at the same pace, and the tooling that wraps these models is where much of the practical difference shows up.
Version Gates and Deprecations: What to Check Before You Switch
Two checks come before any switch. First, deprecation. Grok Code Fast 1 left the supported list on May 15, 2026, so anyone still pointed at that entry has to move, and the changelog names the newer Grok versions and GPT-5 mini as the landing spots. Second, version floors. Grok 4.5 needs Visual Studio 17.14.19, and the floor for Grok 4.6 is not documented yet. If your install predates the required build, the model simply will not appear, and no amount of settings workarounds changes that.
Version gates are not unique to Copilot. When Meta's Muse Spark 13 arrived as a frontier coding model, the rollout was shaped by the same kind of support requirements, and the model's 20 percent reduction in tool calls only mattered on setups that could run it. The lesson transfers across both: check the floor version and the deprecation date before you build a workflow on any model, because the model you chose may quietly disappear from the selector on the next cycle.
What the Documentation Does Not Say Yet
The multi-vendor story has gaps alongside its gains. The minimum IDE version for Grok 4.6 is still TBD in GitHub's documentation, and the supported-models page does not publish side-by-side benchmark numbers, so a like-for-like comparison of vendors on identical code tasks has to be built by hand. What the documentation does give you reliably is status: GA entries, deprecation dates, and version floors, all dated in the changelog. The Copilot documentation hub is where those entries live, and where to re-check as the TBD fields fill in.
If Copilot is in your daily workflow, open the model selector once this week and run the same task on Grok 4.5 and on whatever model you default to. Keep the output that holds up under review. A ten-minute A/B on one task tells you more about which vendor to trust for which part of your code than any changelog entry will.
Conclusion
GitHub Copilot is now a six-vendor platform, with xAI's Grok 4.5 and 4.6 sitting in the model selector as generally available alongside OpenAI, Anthropic, Google, Microsoft, and Moonshot. The practical shift is model per task: pick by task instead of by vendor, and check deprecation dates and version floors before you build a workflow on any of them.
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Which Grok models does GitHub Copilot support?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Grok 4.5 and Grok 4.6, both listed as generally available in GitHub's supported-models documentation for Copilot. The earlier entry, Grok Code Fast 1, was deprecated on May 15, 2026, so it is no longer part of the supported lineup."
}
},
{
"@type": "Question",
"name": "Do I need a specific IDE version to use Grok in Copilot?",
"acceptedAnswer": {
"@type": "Answer",
"text": "For Grok 4.5, yes: the documentation requires Visual Studio 17.14.19 at minimum, and older installs will not show the model in the selector. For Grok 4.6, the minimum version is listed as TBD, so check GitHub's supported-models page for the current requirement before assuming your build qualifies."
}
},
{
"@type": "Question",
"name": "What happened to Grok Code Fast 1 in GitHub Copilot?",
"acceptedAnswer": {
"@type": "Answer",
"text": "GitHub deprecated Grok Code Fast 1 on May 15, 2026, according to the Copilot changelog. Developers on that entry are directed to the newer Grok versions, Grok 4.5 and 4.6, or to GPT-5 mini as an alternative."
}
},
{
"@type": "Question",
"name": "Which vendors does the Copilot model selector cover now?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Six, as of September 2026: OpenAI (GPT-5 series, including mini, Codex, and the 5.6 line), Anthropic (Claude Fable, Haiku, Opus, and Sonnet 5), Google (Gemini, including 3.8 Flash), Moonshot (Kimi K2.7 Code and K3), Microsoft (MAI-Code), and xAI (Grok 4.5 and 4.6)."
}
},
{
"@type": "Question",
"name": "How should I choose between the models in Copilot?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Match the model to the task. For speed and latency, Gemini 3.8 Flash and GPT-5 mini. For complex reasoning, Claude Opus and the GPT-5.6 series. For specialized coding and logic, Grok 4.5, 4.6, and Kimi K3. Because the switch happens mid-workflow in the same editor, you can A/B two models on the same prompt and keep the output that holds up."
}
}
]
}
Frequently asked questions
Which Grok models does GitHub Copilot support?
Grok 4.5 and Grok 4.6, both listed as generally available in GitHub's supported-models documentation for Copilot. The earlier entry, Grok Code Fast 1, was deprecated on May 15, 2026, so it is no longer part of the supported lineup.
Do I need a specific IDE version to use Grok in Copilot?
For Grok 4.5, yes: the documentation requires Visual Studio 17.14.19 at minimum, and older installs will not show the model in the selector. For Grok 4.6, the minimum version is listed as TBD, so check GitHub's supported-models page for the current requirement before assuming your build qualifies.
What happened to Grok Code Fast 1 in GitHub Copilot?
GitHub deprecated Grok Code Fast 1 on May 15, 2026, according to the Copilot changelog. Developers on that entry are directed to the newer Grok versions, Grok 4.5 and 4.6, or to GPT-5 mini as an alternative.
Which vendors does the Copilot model selector cover now?
Six, as of September 2026: OpenAI (GPT-5 series, including mini, Codex, and the 5.6 line), Anthropic (Claude Fable, Haiku, Opus, and Sonnet 5), Google (Gemini, including 3.8 Flash), Moonshot (Kimi K2.7 Code and K3), Microsoft (MAI-Code), and xAI (Grok 4.5 and 4.6).
How should I choose between the models in Copilot?
Match the model to the task. For speed and latency, Gemini 3.8 Flash and GPT-5 mini. For complex reasoning, Claude Opus and the GPT-5.6 series. For specialized coding and logic, Grok 4.5, 4.6, and Kimi K3. Because the switch happens mid-workflow in the same editor, you can A/B two models on the same prompt and keep the output that holds up.
Top comments (0)