Replit makes OpenAI the default while keeping Claude on the menu

Replit now uses OpenAI as the default provider for Replit Agent, while users can still choose Claude. CEO Amjad Masad said performance on an internal benchmark, cost and the quality of the partnership drove the decision.

OpenAI gets the default slot while Anthropic stays available

Masad initially said at TechCrunch Disrupt that Replit was no longer working with Anthropic. He later clarified the statement: the platform remains model agnostic, Claude is still available and OpenAI is the preferred partner for Replit Agent. A person familiar with the companies’ relationship also told TechCrunch that the partnership with Anthropic remained in place.

Replit announced a partnership with OpenAI in August 2026. Before the switch, Masad said Replit had been one of Anthropic’s five largest customers. The default can therefore redirect substantial usage and spending even though the model catalogue remains broader.

A default redistributes demand faster than a comparison table

Most users do not change default settings. For a model provider, winning that slot brings recurring traffic, feedback and revenue without persuading every developer separately. Replit also gains leverage because it can compare models as suppliers to one layer of its own product.

That matters more as business customers grow. Masad said enterprise contracts increased from 2% of revenue last year to about 15% this year, while more than 50% of revenue now comes from B2B. These buyers evaluate price, availability, security and long-term support alongside performance on an individual task.

A vendor benchmark cannot show the full switching cost

Replit says its Sol and Astra agents perform better with OpenAI when building applications from scratch. The public report does not include scores from the internal test or a cost breakdown, so readers cannot assess how large the advantage was. Calling OpenAI the better partner is a commercial judgment, not an independent model evaluation.

Customers are also buying more than a model. They are buying orchestration, error recovery, sandboxing, networking, identity and authorization, which Masad described as the habitat agents need. If that layer absorbs differences between models, a contract may determine the default as readily as a benchmark.

Customer overrides will reveal Claude’s real pull

The clearest signal will be the share of users who manually switch back to Claude, along with production task success and cost after the change. It will also matter whether Replit publishes a comparable benchmark method and whether enterprise customers begin specifying models in their contracts.

Replit expects enterprise and government customers to account for more than half of revenue in 2027. Their requirements will show whether model choice becomes a durable safeguard or merely a setting that few people touch.

Lilith's verdict

OpenAI is now printed on every new Replit user’s ticket. Claude remains on board, but it must wait to see who changes trains for it.