GitHub Copilot now supports multiple LLMs

GitHub is bringing more flexibility and choice to Copilot through the integration of multiple large language models (LLMs).

Since its inception, GitHub Copilot has utilised different LLMs for varied uses. The journey began with the deployment of Codex, an early iteration of OpenAI’s GPT-3, that was fine-tuned specifically for coding tasks. The evolution continued with the launch of Copilot Chat in 2023, initially using GPT-3.5 and subsequently transitioning to GPT-4. As demands evolved, GitHub adapted, employing models from GPT 3.5-turbo to the more recent GPT 4o and 4o-mini, catering to needs for both latency and quality.

“The past year has witnessed a surge in high-quality small and large language models, each excelling in different programming tasks,” stated Thomas Dohme, CEO of GitHub. “It’s clear that the future of AI code generation will be defined not only by multi-model functionality but also by multi-model choice.”

Support for multiple LLMs in GitHub Copilot

In line with this vision, GitHub Copilot is now expanding its capabilities with new models from leading AI developers. Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s latest offerings – o1-preview and o1-mini – augment Copilot’s capabilities.

Initial availability has begun with OpenAI’s o1-preview and o1-mini in Copilot Chat. Anthropic’s Claude 3.5 Sonnet will follow over the next week, with Google’s Gemini 1.5 Pro expected in the coming weeks.

Through these enhancements, GitHub aims to provide developers with greater freedom. Whether coding in VS Code or directly on GitHub.com, individual developers can now select models that best align with their requirements. Organisations and enterprises also have the autonomy to choose which models their teams utilise.

Here’s a detailed breakdown of the new models:

• Anthropic’s Claude 3.5 Sonnet is designed to handle an extensive range of coding tasks, from initial design to maintenance, optimisations, and beyond. Its prowess is evident in managing complex and multi-step coding challenges, making it a robust option for dynamic software development requirements.

• Google’s Gemini 1.5 Pro stands out with a two-million-token context window and its native multi-modal capabilities. This allows it to process code alongside images, audio, video, and text, delivering quick responses for code suggestions, documentation, and explanations.

• OpenAI’s latest models, o1-preview and o1-mini, surpass the capabilities of the current GPT 4o. These models boast enhanced reasoning abilities, providing a deeper understanding of code constraints and edge cases to produce efficient, high-quality results.

With a streamlined login and subscription process, developers can now explore and experiment with different foundational LLMs to find the best fit for their projects.

Introducing GitHub Spark

In addition to the advancements in Copilot, GitHub announced GitHub Spark—a tool that enables the creation of applications in natural language, integrating AI functionalities and external data without the need for cloud resource management.

Sparks provide live previews and foster a creativity feedback loop, enabling users to iterate and refine their developments with ease. This innovation signals a step forward in bringing AI-native tools into mainstream use, helping developers harness the full potential of transformative technology.

With these latest initiatives, GitHub continues to provide developers with versatile tools that help to drive forward its mission as an open platform for innovation and collaboration.

(Image by Gerd Altmann)

See also: JetBrains launches AI model for software development tasks

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

Tags: AI, artificial intelligence, coding, copilot, development, github, github spark, large language models, llm, programming, tools