New Delhi: 

For a platform that was built around offering top AI models, losing direct access to one of the most popular ones overnight has caused real trouble. Developers who relied on Claude within Windsurf’s interface suddenly found themselves scrambling for workarounds, more expensive and less efficient ones. While Anthropic has already rolled out its next-gen Claude Opus 4 and Sonnet 4 models, Windsurf users weren’t given day-one access.

It’s been a bumpy week for Windsurf users. The Vibe Coding startup, recently acquired by OpenAI for around ₹26,100 crore (USD 3 billion), has run into sudden restrictions from Anthropic. The , especially Claude 3.5 Sonnet and 3.7 Sonnet, were pulled with less than five days’ notice, according to Windsurf CEO Varun Mohan.

What exactly happened?

Varun Mohan wrote on X that Anthropic “cut off nearly all of our first-party capacity to all Claude 3.x models.” Windsurf says it tried to keep those models available, even offering to pay for full capacity, but Anthropic didn’t budge. Now, Claude 3.5 and 3.7 are only available to Windsurf users through “bring-your-own-key” setups, not directly from the platform.

The blog post confirmed, “With less than a week of notice, Anthropic informed us they were cutting off nearly all of our first-party capacity to the Claude 3.x models.”

The timing of this move couldn’t be worse. Windsurf just became part of the OpenAI fold, and the platform had built a loyal dev base around its “vibe coding” tools. With tools like Gemini 2.5 Pro and GPT-4.1 still running strong, Windsurf is now trying to steer users toward these alternatives. They’ve even dropped the Gemini 2.5 Pro rate to 0.75x to help soften the blow.

Why it matters for devs

Mohan has pointed out, “We may see some short-term Claude 3.x model availability issues as we have very quickly ramped up capacity on other inference providers.” The company was already working to find alternatives.

But the real tension is about access and fairness. While Windsurf was forced to scramble, other coding platforms like Cursor, Devin, and GitHub Copilot all had access to Claude 4 models at launch.

The Claude models were known for good reasoning and code-writing accuracy. Claud Opus 4 scored 72.5% on SWE-bench and 43.2% on Terminal-bench. Sonnet 4 even outperformed it slightly with 72.7%. These models were central to what made Windsurf stand out.

Windsurf isn’t giving up

The team made it clear the situation isn’t ideal, but they’re not folding. “The magic of Windsurf has never been limited to the model,” the blog stated. “The magic is in the deep contextual understanding… in the intentional and thought-out UX… in the customizations like Workflows, Rules, and Memories.”

Still, for developers relying on Windsurf for Claude, this is a sign of rising tensions between AI labs and the platforms that depend on them.

Read More:

CEO says insider wiped GitHub and AWS servers