Replit has made OpenAI the default provider for Replit Agent while retaining Anthropic’s Claude as an available customer choice. CEO Amjad Masad clarified the arrangement after initially saying that Replit was no longer working with Anthropic, according to TechCrunch.
The new fact is narrower than a complete supplier departure. OpenAI receives the preferred position in Replit’s app-building product, but the platform continues to support models from Anthropic and other providers. For teams using Replit to develop internal tools or prototypes, the operational question is therefore which model runs by default—not whether Claude has disappeared.
The change is a default, not a removal
Replit provides a cloud development environment in which users can build and deploy applications from a browser. Masad told TechCrunch that the company remains model agnostic, that customers can choose Claude or other supported models, and that OpenAI is now the preferred partner behind Replit Agent.
This clarification separates two decisions that can otherwise appear identical. Replit has selected a preferred supplier for the standard Agent experience, while preserving alternative models within the platform. The source does not specify whether every workspace, project type or existing customer receives the same default, nor does it identify the exact OpenAI model used in each workflow.
That missing detail matters for comparison. A team cannot infer from the provider announcement alone that an existing project has changed models, or that selecting Claude produces the same sequence of actions as the default Agent configuration. The documented conclusion is limited to provider preference and continued model availability.
Defaults shape routine model use
A default influences the path that requires no additional choice. TechCrunch notes that customers commonly remain with the preset option and reports that the switch could reduce Replit’s spending with Anthropic. The article does not disclose Replit’s supplier costs or quantify the expected shift in model usage.
For commercial operations teams building a quoting tool, lead-routing utility or another internal application, this changes the starting point for evaluation. A new build initiated through the standard Replit Agent experience will encounter OpenAI as the preferred provider. Claude remains relevant when a user deliberately selects it, where that choice is offered.
The provider setting does not establish which model will produce a more accurate application for a particular business process. It also does not prove that projects will become cheaper, faster or easier to maintain. Those outcomes depend on the task, the generated application and Replit’s implementation around the underlying model.
Conceptual illustration, not a photograph, showing how a default route differs from an explicit model choice.
Replit cites app building, economics and partnership
Masad attributed the preference to performance, cost and the strength of Replit’s relationship with OpenAI. He also said Replit uses an internal benchmark and publishes a public benchmark, with its Sol and Astra systems performing better at building applications from scratch, as reported by TechCrunch.
These are company-reported grounds for the decision, not an independent comparison supplied in the article. The source does not provide the benchmark prompts, model configurations, review criteria or results needed to determine how the evaluation maps to a specific commercial application.
The task definition is still useful. Replit is describing performance in end-to-end application creation rather than presenting a general claim that one provider is superior across coding work. A buyer assessing Replit for browser-based app development should therefore compare the complete build outcome: whether the requested workflow is represented correctly, whether revisions preserve intended behaviour and whether the deployed application performs the required job.
Conceptual illustration, not a photograph, of the factors Replit says informed its provider preference.
The supplier relationship remains disputed
Masad’s clarification resolves part of the apparent contradiction in his original statement: Claude remains accessible even though OpenAI has become preferred. It does not fully settle the status of the commercial relationship. An unnamed person familiar with the companies told TechCrunch that the Anthropic partnership was still in place.
The disagreement is about how to describe the supplier relationship, not about the customer-facing availability confirmed by Replit’s CEO. Our earlier coverage of the conflicting accounts surrounding Replit and Anthropic examines that distinction in more detail.
For teams choosing a model inside Replit, the immediate decision is more concrete than the partnership label. Confirm which provider is selected for the relevant Agent workflow, verify that Claude remains selectable where required, and compare both options on the application the team intends to build. The next evidence to watch is Replit documentation identifying the exact default models, the workflows covered by the change and the methodology behind its app-building benchmark.
