GitHub has just announced the release of a powerful new code completion model for GitHub Copilot called GPT-4o Copilot. This model brings significant improvements in both the speed and quality of AI-assisted code completions.
A Brief History of GitHub Copilot
GitHub Copilot first launched in technical preview back in June 2021 as a collaboration between GitHub and OpenAI. It leveraged the GPT-3 language model, which had been trained on a huge corpus of public code, to offer intelligent code completion suggestions directly in the IDE.
Over the past two years, Copilot has evolved rapidly. It expanded support for more languages, improved its understanding of code context and conventions, and became available in popular IDEs like Visual Studio Code and JetBrains.
In 2022, GitHub Copilot for Individuals was released, making the tool available to all developers for a monthly subscription. Then in 2024, GitHub made Copilot free for all developers.
Throughout this journey, the AI models powering Copilot have continued to get smarter and more sophisticated. The new GPT-4o Copilot model represents the latest leap forward.
What's New in GPT-4o Copilot?
The GPT-4o Copilot model enhances the core "ghost text" code completion capabilities of GitHub Copilot. Some of the key improvements include:
Faster Completions: Suggestions appear nearly instantly as you type, without needing to hit enter to trigger the AI.
Improved Accuracy: The model was trained on over 275,000 high-quality public code repositories, enabling it to provide more relevant and accurate code suggestions.
Expanded Language Support: While many AI code assistants focus on just a few popular languages, GPT-4o Copilot was trained on over 30 programming languages. This includes Python, JavaScript, TypeScript, Ruby, Go, C++, C#, Java, PHP, Swift, Rust, and many more.
Optimized for Real-World Code: By learning from a curated set of public codebases, GPT-4o Copilot has a strong understanding of clean, idiomatic, real-world programming patterns. It can help you write code that follows industry best practices.
Under the hood, GPT-4o Copilot is built on OpenAI's GPT-4o mini architecture. However, GitHub has fine-tuned and optimized the model specifically for the task of code completion using its vast corpus of public source code.
Getting Started with GPT-4o Copilot
The new code completion model is available starting today for GitHub Copilot users in Visual Studio Code. If you have Copilot installed, you can enable it by:
- Clicking the Copilot icon in the VS Code title bar
- Selecting "Configure Completions" and then "Change Completions Model"
- Choose "gpt-4o-copilot" from the model options
Once enabled, Copilot will immediately start using the new model as you code. Completions will appear inline as you type. Pressing Tab will accept the suggestion.
GPT-4o Copilot will be rolling out soon to Copilot users in JetBrains IDEs as well. GitHub plans to make it the default model for all Copilot users in the near future after monitoring its performance and gathering user feedback.
The Future of AI-Assisted Coding
From its early technical preview to becoming an essential tool for developers around the world, GitHub Copilot has come a long way in a short time. It has already revolutionized the way many developers write code.
With the GPT-4o Copilot update, GitHub is pushing the boundaries of what's possible with AI code completion even further. As these language models continue to evolve, they have the potential to make software development faster, easier, and more accessible than ever before.
To learn more and start using the GPT-4o Copilot model, visit the GitHub Copilot website. Happy coding!
Top comments (2)
Thanks for sharing, Andrea!!
love this!