GitHub Copilot added four more frontier AI models in a single week — the real problem now is choosing, not access
GitHub Copilot's model picker gained Grok 4.7, GPT-6 Sol, GPT-6 Luna and Claude Opus 5.5 within days of each other (21–22 September 2026), on top of Claude Fable 5.1 and GPT-6 Astra added two weeks earlier — meaning teams now choose between eight-plus frontier AI coding models with no clear default, which is a process problem more than a technology one.
26 September 2026
GitHub shipped four model additions to Copilot’s picker inside a single week: Grok 4.7 went live on 21 September 2026, followed the next day by GPT-6 Sol, GPT-6 Luna and Claude Opus 5.5. That’s on top of Claude Fable 5.1 and GPT-6 Astra, both added in the first week of September. Copilot’s model picker — available across VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, the CLI, the coding agent, GitHub Mobile and github.com — now spans OpenAI, Anthropic and xAI models side by side, each pitched at a slightly different job: GPT-6 Sol for balanced multistep validation, GPT-6 Luna as the cheap, fast option, Grok 4.7 with a cheaper output but pricier cached input, and Opus 5.5 for long-running agentic work and knowledge tasks.
Model breadth isn’t the constraint anymore
For most of 2026, the AI coding story was about capability catching up to hype. That’s no longer the gating issue — access to frontier models is now abundant and largely commoditised across the major IDEs. What’s replaced it is a genuinely harder problem: with eight-plus models available and each priced and tuned differently, “which model should this task use” has become a decision that needs an actual policy, not a developer’s personal preference in the moment. Left undecided, teams either default to whatever loads first — often not the best fit for the task — or burn time model-shopping on every ticket.
So what
A development team that treats model selection as a deliberate, task-by-task decision — matching model cost and strengths to what a piece of work actually needs — gets a materially different cost and quality outcome than one that picks once and never revisits it. This is exactly the kind of operational discipline that separates a considered AI-assisted build from one that’s just pointed at whatever’s newest. See our AI-assisted development approach for how we make and revisit these calls through a project, or get in touch if your current build is running on a model choice nobody’s checked in months.