The biggest partnership breakup in AI developer tools didn’t happen on a stage. It happened quietly, two months ago, in a settings menu.
Replit CEO Amjad Masad told TechCrunch onstage at Disrupt this week that the coding platform is no longer working with Anthropic, the company whose Claude models once powered it. When Replit announced its OpenAI partnership in August, it framed the move as an expansion. It never mentioned that Anthropic was getting pushed out.
“I think ChatGPT is better now,” Masad said, pointing to an internal benchmark called VBench and a public benchmark the company publishes. He claimed OpenAI’s Sol and Astra models consistently outperform the competition at building vibe-coded applications from scratch, and said the switch came down to “cost, partnership, and performance.”
The Default Is the Deal
Here is the part everyone is glossing over: Anthropic wasn’t actually deleted from Replit. In a follow-up, Masad clarified that customers can still access Claude on the platform. What changed is the default. OpenAI’s ChatGPT is now the default model powering Replit Agent, and most users never touch the settings menu.
That’s the real economic event. The Tech Buzz notes that the shift could significantly reduce Replit’s spending with Anthropic, since defaults decide where the API calls — and the API bills — actually go. Replit had been, until recently, one of Anthropic’s top five customers. Losing the default slot in a top-five account is worth more than any press release about “model agnosticism.”
“Replit remains model agnostic,” Masad insisted, adding that “OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.” Model agnostic, preferred partner — the wording tells you everything about which one actually matters.
He Said, They Said
There is genuine messiness under the messaging. People familiar with the partnership told TechCrunch the deal with Anthropic is still in place, and Replit’s own website and blog were promoting Anthropic’s models as recently as late September — one post cited Claude Fable 5 and 5.1 as running in Replit Agent’s production environment, while a product employee was hyping Opus 5.5 on X around the same time.
So which is it: breakup, or not? The honest answer is that the binary is the wrong frame. A platform can be “model agnostic” on the website and still route nearly all of its spend to one vendor through the default. The transition can be months old before the CEO says a word onstage. That ambiguity is the strategy.
Why this matters
This is the clearest signal yet that the AI platform wars have entered the cost era. When Masad lists “cost” first among his reasons, he is describing a market where model differentiation has narrowed enough that pricing and partnership terms decide deals. Replit’s unit economics run on API costs; OpenAI’s aggressive pricing for platform deals gave it room to grow without blowing up the cost structure.
Watch for the playbook to spread. Replit’s quiet transition — announce the new partner as an “expansion,” never mention the old one’s exit, let the default do the switching — is a template other cost-squeezed platforms can copy. And for Anthropic, the risk isn’t one lost customer; it’s the gravitational pull OpenAI is building around developer defaults. The models that power a generation of coders’ daily workflow earn a kind of mindshare no benchmark can measure.
Masad, meanwhile, is steering Replit toward enterprise: enterprise revenue went from 2 percent last year to about 15 percent this year, with more than half of revenue now B2B. He expects enterprise and government contracts to clear half of revenue in 2027. When your buyers are governments, the model’s default setting — and who controls it — stops being a product detail and becomes procurement strategy.
Meanwhile in Apps & Tools: Google is letting you keep your vibe-coded apps private on the Play Store, and OpenAI’s Codex cloud explained.
