Replit makes OpenAI its default, dropping Claude as the model behind its agent

Replit built much of its coding product on Anthropic's models, and until recently it was one of Anthropic's five largest customers. At TechCrunch Disrupt, CEO Amjad Masad said that is over: OpenAI is now the preferred provider behind Replit Agent, and ChatGPT-based models are the default. He framed the call around three things, cost, partnership, and performance, and said OpenAI's models consistently do better at building vibe-coded applications from scratch on both an internal benchmark and a public one.

The story is not entirely clean. An insider told TechCrunch the Anthropic relationship is still in place, and as recently as September Replit's own blog was promoting Claude. Masad stressed that Replit stays "model agnostic" and that Claude remains available to anyone who selects it. But defaults decide almost everything, since most users never change them, so Replit's spending with Anthropic could fall sharply.

Underneath the model switch is a business shift. Masad said enterprise went from about 2 percent of revenue last year to roughly 15 percent this year, more than half of revenue is now B2B, and he expects enterprise and government to pass half in 2027. Picking the model that wins his internal coding benchmark matters more when the buyers are companies.

Why it matters

If you build on Replit, the model generating your app changed by default, so test whether your prompts and results hold on the new backend. For everyone else it is a concrete data point on OpenAI gaining ground in exactly the coding workloads where Anthropic has led.