CEO Amjad Masad cited cost, performance, and partnership considerations, and said Replit uses an internal benchmark. This is a default-routing change, not a platform-wide removal of alternatives: Masad told TechCrunch customers can still select Claude and other supported models.
Thesis
For production agents, the harness is becoming a more durable product boundary than the identity of the model running inside it.
Observed facts
Replit made OpenAI’s ChatGPT its preferred/default provider for Replit Agent, according to CEO Amjad Masad.
1Masad said customers can still choose Anthropic’s Claude and other supported models.
1Masad cited cost, performance, and partnership considerations and referred to an internal benchmark.
1TechCrunch reported Replit had been a top-five Anthropic customer and that recent company materials referenced Anthropic models in production.
1
Mechanism
Changing a product default shifts likely usage and provider spend without removing alternative models from the platform.
Why now
Replit’s reported switch makes model choice a live product and procurement decision, while retaining alternatives inside the same agent platform.
Implications
Teams evaluating Replit Agent should distinguish the default model from models users can explicitly choose, since defaults can influence behavior and provider economics.
Evidence boundary
The source reports Replit’s stated rationale and the provider/default arrangement, but offers no independently comparable benchmark results or quantified usage shift.
Watch conditions
Watch whether Replit publishes benchmark methodology or usage data showing how the new default changes model selection and outcomes.