GitHub’s Copilot Chat is now multimodal and no longer relies on OpenAI

GitHub announced this week a major update for users of its AI tool Copilot Chat. From now on, developers will be able to choose from a wide variety of large language models beyond GPT-4 from OpenAI, which was the only one available until now.

Subscribe to the Softonic newsletter and get the latest in tech, gaming, entertainment and deals right in your inbox.

Subscribe (it's FREE) ►

The platform will allow users to choose models such as Claude 3.5 Sonnet from Anthropic, Gemini 1.5 Pro from Google, and the latest models from OpenAI: o1-preview and o1-mini. Anthropic will be the first model to be added, followed by Google’s Gemini in a few weeks.

Thomas Dohmke, CEO of GitHub, justified to TechCrunch this expansion in access to different models as a way to offer more options to developers, thus avoiding an exclusive dependency on OpenAI. Although some analysts see this as a maneuver by Microsoft to reduce that dependency, Dohmke prefers to frame it from the perspective of flexibility, highlighting that “not all models excel at all tasks.” By offering various models, GitHub responds to the different needs of developers, who may require specific models for certain languages or tasks.

“We truly believe that the era of a single model is over,” stated Dohmke. The CEO also explained that GitHub Copilot already used different versions of GPT-3 and GPT-4 depending on the latency and accuracy required. The new offering with Anthropic and Google reflects GitHub’s interest in “developer choice,” allowing them to use models that already fit their preferences or are integrated into other projects.

This possibility of choosing between different models was already in the testing phase when GitHub enabled the choice between GPT-4 and GPT-4 o1 a month ago. For now, this new option is limited to Copilot Chat and the recent Spark tool, although Dohmke assured that it is expected to be extended to more tools in the future.

GitHub DOWNLOAD