GitHub has expanded Copilot’s model support to include new Anthropic, Google, and OpenAI models, and introduced GitHub Spark, an AI-driven tool that builds web apps from natural-language prompts.
Developers using GitHub Copilot, the AI-powered coding assistant, can now choose from multiple generative AI models. The newly supported models include Google Gemini 1.5 Pro, OpenAI GPT-4o, OpenAI o1-preview, OpenAI o1-mini, and Anthropic Claude 3.5 Sonnet. Initially, multi-model support is available in GitHub Copilot Chat, with plans to extend it across other Copilot surfaces and features soon.
GitHub announced the expanded model list on October 29 and simultaneously unveiled GitHub Spark, a native-AI tool designed to let users create fully functional web applications using plain language.
Developers using Copilot on github.com or in Visual Studio Code can switch between models during a Copilot Chat session to pick the best model for each task, or let Copilot continue using its powerful default model. The multi-model approach enables developers to work with a range of leading models within familiar workflows. Anthropic Claude 3.5 Sonnet is available in GitHub Copilot today, Google Gemini 1.5 Pro will be rolled out in the coming weeks, and the new OpenAI models are available now. GitHub said it will keep collaborating with model providers to expand multi-model choices across more Copilot experiences.
GitHub Spark aims to simplify app creation for developers of all skill levels by letting them describe ideas in natural language to build micro-apps called “sparks.” Sparks are lightweight, fully functional micro-apps that can incorporate AI features and connect to external data sources without requiring developers to manage cloud infrastructure. Using an iterative feedback loop, users start with an initial prompt that can leverage both OpenAI and Anthropic models, view live previews as the app is assembled, and save versions of each iteration to compare changes. A spark can run on desktop, tablet, or mobile devices. GitHub plans to iterate on Spark to make it as intuitive as possible for both non-technical users and professional developers.
GitHub also announced substantial updates to Copilot for use in the Visual Studio Code editor, GitHub Copilot Workspace, GitHub Models for building generative AI applications, and GitHub Copilot Autofix code scanning. Users of GitHub Copilot Chat in VS Code can now apply changes across multiple files at once. The new multi-file editing mode, available from November 1, enables complex project-wide edits driven by natural-language prompts. GitHub Copilot Autofix now includes security campaigns to help developers and security teams remediate vulnerabilities at scale, with the ability to triage as many as 1,000 alerts at a time.
Finally, GitHub said Copilot Extensions will become generally available in early 2025. Currently in preview, GitHub Copilot Extensions let developers ask questions about any Copilot-integrated developer tools or services, including offerings like Atlassian, Docker, Sentry, and Stack Overflow. The company also noted that Copilot’s code-completion features are now available in a public preview for Apple’s Xcode IDE.