GitHub Launches Copilot Enterprise for Enhanced Code Completion and Team Collaboration
GitHub has officially launched Copilot Enterprise, its advanced code completion tool designed specifically for larger businesses, at a price of $39 per month. This new offering builds upon the existing Business plan, which includes IP indemnity, while introducing several key features tailored for bigger teams. A standout capability is the ability to access and reference an organization’s internal code and knowledge base. Notably, Copilot is now integrated with Microsoft’s Bing search engine (currently in beta), and users will soon have the option to fine-tune Copilot’s models based on their organization’s existing codebase.
This means that new developers can query Copilot for specific tasks, such as deploying a container image to the cloud, receiving answers tailored to their organization’s unique processes. For many developers, navigating different workflows—not just understanding the code—is often a significant challenge when transitioning between companies, although Copilot can assist with code comprehension as well.
Many teams already store their documentation in GitHub repositories, making it easier for Copilot to understand and utilize this information effectively. GitHub CEO Thomas Dohmke highlighted that, since the company keeps most of its internal documents within GitHub, employees have begun leveraging these features for non-technical inquiries as well—such as questions about vacation policies.
Dohmke explained that customers have expressed a need for internal information references since the inception of Copilot. “Developers often engage with processes and libraries that differ from their personal or open-source projects. Many organizations have unique tools, systems, and dependencies that don’t exist outside of their ecosystem,” he stated.
Regarding the Bing integration, Dohmke emphasized its importance for retrieving up-to-date information that may have changed since the original training of the model—particularly regarding open-source libraries or APIs. Currently, this feature is exclusive to the Enterprise version, though there’s potential for its introduction across other subscription tiers in the future.
One element likely to remain an Enterprise-exclusive feature due to its associated costs is the fine-tuning capability, set to launch soon. “We empower organizations to select their repositories on GitHub and refine the model based on those specific resources,” Dohmke explained. “This generation of an optimized model, customized for their environment, will enhance Copilot’s performance.” He also noted that while this offering allows for tailored models, it may not provide updates as quickly as other methods involving embeddings and agents (like the new Bing agent). However, clients currently testing this feature have reported significant improvements, especially those working with less common programming languages or proprietary internal libraries.
Beyond the launch, I inquired about the broader vision for Copilot going forward. The response was clear: “more Copilot in more contexts.” Over the coming year, the focus will increasingly emphasize integrating Copilot directly into existing workflows, rather than requiring users to shift to a separate tool for content creation. “At GitHub, we are thrilled about the potential of embedding Copilot directly into github.com, where developers are already collaborating and crafting the future of software.”
In terms of technical infrastructure, Dohmke mentioned that the auto-completion feature currently operates on GPT-3.5 Turbo. Due to latency considerations, GitHub has not transitioned to GPT-4, although the team has updated its model numerous times since launching Copilot Business. Unlike Google, GitHub doesn’t plan to tier its pricing based on model size. “Different applications necessitate distinct models. Factors like latency, accuracy, and ethical AI practices play a significant role in ensuring we deliver high-quality code that meets our customers' standards. We’ll continue to optimize Copilot by utilizing the best models suited for various scenarios,” Dohmke affirmed.
Keywords: GitHub, Copilot Enterprise, code completion tool, developer collaboration, Microsoft Bing integration, AI, fine-tuning models
Ranking
