GitHub Copilot Starts Broad GPT-5.4 Rollout Across IDEs, CLI, Mobile, and the Web

GitHub said on March 5, 2026 that GPT-5.4 has reached general availability inside GitHub Copilot. The company describes GPT-5.4 as OpenAI’s latest agentic coding model and says its early testing showed stronger real-world performance on intricate, multi-step, tool-dependent software tasks.

The bigger story is distribution. GitHub is not confining the model to one editor or one premium tier. Instead, it is making GPT-5.4 selectable across most of the surfaces where Copilot already operates. That means a developer can move between local IDE work, command-line flows, web sessions, and mobile review without switching away from the same model family.

Where GPT-5.4 is available

• Visual Studio Code v1.104.1 and later in chat, ask, edit, and agent modes.
• Visual Studio 17.14.19 and later, JetBrains 1.5.66 and later, Xcode 0.48.0 and later, and Eclipse 0.15.1 and later.
• github.com, GitHub Mobile on iOS and Android, GitHub CLI, and the GitHub Copilot Coding Agent.
• Copilot Business and Enterprise customers need administrators to enable the GPT-5.4 policy in Copilot settings.

That footprint matters because model selection is increasingly part of the product, not just infrastructure hidden behind it. By spreading GPT-5.4 across editors, the web, mobile, and CLI, GitHub is positioning Copilot less as an editor add-on and more as a unified developer environment with shared model access. It also makes the competition between coding assistants more about workflow coverage and policy control, not just benchmark claims.

For developers, the practical implication is broader access to a newer reasoning-oriented coding model in the places where actual work happens. For enterprises, the admin-controlled enablement path gives teams a way to phase adoption while preserving governance. That combination of broad rollout plus policy gating is becoming a common pattern for enterprise AI deployment.

Source: GitHub changelog