GitHub said on July 5 that Anthropic’s Claude Sonnet 5 is now available inside Copilot, its AI coding tool — the first time a model from outside the Microsoft ecosystem has been offered on the platform. The change, effective immediately, lets developers choose between OpenAI’s Codex and Anthropic’s Claude in the same interface.
The announcement is small in mechanics and large in meaning. For years, Copilot ran on a single model supplier: OpenAI, in which Microsoft holds a major stake. The move to multi-model is a strategic shift, acknowledging that developers want choice at the point of work, not at the point of procurement. A developer who prefers Claude’s style on reasoning tasks and Codex on routine code can now switch without leaving the editor.
The mechanics are straightforward. Sonnet 5 appears as a selectable model in Copilot’s model picker, alongside the OpenAI models already there. GitHub said the Anthropic model supports the same workflows as the existing options — code completion, chat, agent mode — and that pricing follows the standard Copilot subscription structure. No separate signup is required, a decision that reflects how the two companies want the integration to feel: as a feature, not a partnership announcement.
The timing reflects the state of the market. Coding assistants have become the most competitive corner of AI, with every major lab selling one and few developers loyal to a single brand. GitHub, which hosts much of the world’s code, is in a position to be the neutral floor where models compete on merit. That neutrality, once a side effect of its market share, is now a product strategy: if all the models live in Copilot, GitHub captures the usage regardless of which one wins the technical contest.
The decision also reshapes the relationship between Microsoft and OpenAI. The two companies remain deeply connected — OpenAI’s models power Microsoft’s consumer products, and Microsoft is a major investor — but the exclusive alignment has loosened over the past year. Offering Claude inside Copilot does not break the partnership; it renegotiates its terms, making OpenAI one supplier among several rather than the default.
The shift also mirrors a broader change in how Microsoft sells AI. The company has moved from bundling OpenAI’s models into everything it ships to positioning itself as a multi-model cloud, where customers pick the frontier lab they want. The Copilot decision applies that logic to the product developers touch every day. People close to the company said the calculation was simple: developers were already using Claude alongside Copilot through workarounds, and offering it directly converts that shadow usage into measured, billable usage.
For Anthropic, the deal is access to distribution it could not build alone. Claude has a loyal following among developers, but reaching them inside Copilot’s interface puts the model in front of a much larger population, including the non-engineers who now use coding tools for automation. Sonnet 5’s strengths in long-context reasoning and instruction-following are a natural fit for the assistant use case, and the timing — days after the model’s pricing controversy — gives Anthropic a chance to refocus attention on what the model does.
The integration lands days after Anthropic acknowledged that its Sonnet 5 pricing was not a real discount, an episode that frustrated some developers. Distribution deals do not fix pricing, but they change the context in which the question gets asked: a developer choosing between models inside Copilot is comparing output, not launch messaging. For Anthropic, the timing is fortuitous; for its rivals, it shows that the market for coding models is now decided in the interface, not the press release.
Developers who tested the integration over the weekend described the practical appeal: same prompt, two models, immediate comparison. The ability to run the same task through both systems, side by side, is precisely the kind of transparency that enterprise buyers have been demanding, according to people who advise companies on AI procurement. It also raises the pressure on both labs to keep their models sharp, since switching costs inside Copilot are close to zero.
The move also puts pressure on the model providers themselves. OpenAI has treated Copilot as its distribution channel into Microsoft’s developer base, and the arrival of Claude changes the terms of that relationship: Codex now competes inside its own home turf for the same prompts. For Anthropic, the integration is a rare win in a market where it has trailed OpenAI in raw usage. Both companies will be watching the same number — the share of Copilot sessions that route to each model — and that number will now be visible to the public in a way that internal adoption metrics never were.
The longer-term question is whether the platform model wins. If Copilot becomes the place where all frontier models are available, Microsoft collects the toll on AI-assisted software development while the labs fight over quality. If developers instead route around platforms and call models directly, the integration is one more tab in an increasingly crowded editor. For now, the announcement does what Microsoft intends: it makes Copilot the obvious place to compare the two most prominent models in coding — and keeps GitHub at the center of the conversation.
