Claude 3.5 Sonnet lands on GitHub Copilot

For the past few months, plenty of developers have been quietly jumping ship from GitHub Copilot to newer tools like Cursor. The reason was simple: Anthropic's Claude 3.5 Sonnet had become the undeniable king of code, and Copilot was strictly an OpenAI shop.

That exclusivity officially ended today. At its annual Universe conference, GitHub announced that Copilot is going multi-model. Developers will soon be able to swap between models from Anthropic, Google, and OpenAI directly inside VS Code and GitHub.com.

The model buffet arrives in VS Code

The marquee addition rolling out today in public preview is Claude 3.5 Sonnet. Copilot Chat users and organizations will see it land over the coming weeks, running through Amazon Bedrock cross-region inference to keep latency and uptime under control.

Anthropic made sure to flaunt its recent benchmark wins alongside the news. The upgraded Claude 3.5 Sonnet holds the top spot on SWE-bench Verified for solving real-world GitHub issues and boasts a 93.7% score on the HumanEval Python benchmark.

Google's Gemini 1.5 Pro will follow in the coming weeks, while OpenAI's lineup is expanding to include GPT-4o alongside the reasoning-heavy o1-preview and o1-mini. Developers can actively toggle between these models mid-chat, matching specific debugging or drafting tasks to the engine best suited for the job.

As GitHub CEO Thomas Dohmke put it, no single model rules every scenario anymore, and developers want the agency to pick their own tools.

Beyond switching brains

GitHub didn't stop at model selection. The company also unveiled Spark, an experimental AI tool designed to build shareable web micro-apps entirely from natural language prompts. It mixes OpenAI and Anthropic models under the hood, provides live visual previews, and lets novices build software without touching raw syntax.

For power users, VS Code gets multi-file editing on November 1st, allowing Copilot Chat to coordinate edits across multiple project files at once. Apple developers are also getting attention with Copilot for Xcode entering public preview, alongside a brand-new automated code review feature and an extensions platform slated for early 2025.

Why it matters

This is a massive strategic pivot. GitHub Copilot was the very first poster child of Microsoft's multibillion-dollar bet on OpenAI back in 2021. Watching GitHub open its front door to chief rivals Anthropic and Google feels like a public admission that no single AI lab can monopolize developer tooling.

For developers, this is an immediate win. You no longer have to ditch your company-approved enterprise setup or pay for separate editor subscriptions just to tap Claude's superior code comprehension. It eliminates the friction of copying context between web interfaces and the terminal.

Strategically, GitHub made the smartest move available. Over 1.8 million paying subscribers and 100 million registered developers live in GitHub's ecosystem. By uncoupling its interface from OpenAI's exclusive roadmap, GitHub protects its distribution moat. It turns Copilot from a proprietary wrapper into an open platform where frontier model labs must compete for developer attention.

The only real question mark now is unit economics. High-end reasoning models like o1 and Anthropic's Sonnet are notoriously compute-heavy. Copilot's flat subscription pricing might soon feel the squeeze if power users start hammering Claude all day long.