GitHub Expands Copilot with Anthropic, Google, and OpenAI Models, Empowering Developer Choice
San Francisco, CA – At GitHub Universe, GitHub announced the integration of multiple new large language models (LLMs) into GitHub Copilot, including Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s o1-preview and o1-mini. This move marks a new chapter in developer choice, allowing users to select the best AI models for their needs.
GitHub Copilot has leveraged different LLMs for various use cases since its inception. Initially built on Codex, an early version of OpenAI's GPT-3 tailored for coding, Copilot has evolved through several model upgrades, including GPT-4. Today, GitHub is taking a major step forward by offering developers the choice between multiple LLMs to suit specific coding needs.
- Claude 3.5 Sonnet: Specialized in software lifecycle management and complex coding tasks
- Gemini 1.5 Pro: Features 2M token context window and multi-modal processing
- OpenAI o1 Models: Advanced reasoning capabilities surpassing GPT-4o
- Developer choice in selecting optimal AI models
- Organizational control over model availability
- Enhanced code quality through specialized model capabilities
- Streamlined application development with GitHub Spark
New Models, Greater Choice
The expanded capabilities will first roll out in Copilot Chat, with OpenAI’s o1-preview and o1-mini models available now. Anthropic’s Claude 3.5 Sonnet will be introduced progressively over the coming week, while Google’s Gemini 1.5 Pro will follow in the next few weeks. These models will be integrated into a range of Copilot features, from multi-file editing to code review and security autofix.
This multi-model approach empowers individual developers to select the LLM that best fits their projects, while organizations have complete control over which models they make available for their teams. "GitHub is committed to ensuring every developer has the agency to build with the models that work best for them," said Thomas Dohmke, CEO of GitHub.
Meet the New Models
- Anthropic’s Claude 3.5 Sonnet excels across the software development lifecycle—from design and bug fixes to maintenance and optimization. It handles complex, multi-step coding tasks and is well-suited for scenarios like legacy app updates and feature development.
- Google’s Gemini 1.5 Pro boasts a two-million-token context window and multi-modal capabilities, meaning it can process code, images, audio, video, and text simultaneously. Its versatility and impressive response times make it ideal for everything from regular code suggestions to documentation and code explanation.
- OpenAI’s o1-preview and o1-mini introduce advanced reasoning capabilities that surpass GPT-4o. These models provide a deeper understanding of code constraints and edge cases, enhancing efficiency and quality in generated outputs.
Introducing GitHub Spark
As part of its vision to reach one billion developers, GitHub also unveiled GitHub Spark—an AI-native tool for building applications entirely using natural language. These micro apps, called Sparks, integrate AI features and external data sources without requiring cloud resource management. Users can provide initial prompts and see live previews of their apps as they develop, creating a seamless and intuitive experience for app building.
"GitHub Spark is designed to accelerate the journey of creating innovative software, from initial ideas to complete applications, all with the power of AI at your fingertips,” said Dohmke.
