Replit is no longer working with Anthropic, CEO says

At TechCrunch Disrupt, Replit CEO and co-founder Amjad Masad said the AI coding platform is no longer working with Anthropic. He pointed to cost, partnership, and performance, saying ChatGPT is better now and that internal and public benchmarks show Sol and Astra performing better on building vibe-coded apps from scratch. Replit announced an OpenAI partnership in August; there had been no prior signal that Anthropic was out.

People familiar with the Anthropic–Replit partnership disputed Masad’s framing, saying the partnership was still in place. TechCrunch notes Replit blog and social posts from late September still citing Claude Fable 5 / 5.1 in Replit Agent’s production environment and Opus 5.5. Masad also said enterprise is now about 15% of revenue, up from roughly 2% last year, with more than 50% of revenue B2B.

Model switches at coding platforms are product news and vendor politics at once. If Replit really is consolidating on OpenAI’s stack, that’s a public vote on price and “vibe coding” benchmarks — and a hit to the narrative that every serious coding agent needs a Claude seat.

We’re not treating the stage line as settled fact. TechCrunch’s sourcing says the partnership may still be live, and Replit’s own recent posts still name Anthropic models. Until Replit clarifies, the honest desk read is: Masad wanted the room to hear OpenAI is winning the bake-off; the paperwork may lag the rhetoric.

Either way, the pattern matters. Coding agents are sticky products with thin margins once you pay frontier rates at scale. Cost + partnership + benchmark is exactly how those deals get rewritten. I’m watching for a clean statement from Replit or Anthropic, and whether Claude disappears from Agent’s production docs.

Replit’s Anthropic partnership had been credited with helping revenue growth. Masad also talked about agent “habitat” — orchestration, error recovery, sandboxing, identity — as the real safety layer for deploying agents.