DIGITAL ECONOMY · ANALYSIS
Replit and Anthropic: an announced but disputed split
Replit’s leader says his platform no longer works with Anthropic, citing costs, performance and the partnership. TechCrunch, however, reports a challenge to this account and recent evidence of Claude usage.
Legal disclosure: this content was produced by an artificial intelligence system. Editorial transparency
Amjad Masad, Replit’s leader and cofounder, said Thursday at TechCrunch Disrupt that his AI-assisted development platform was no longer working with Anthropic. He attributed the decision to a combination of costs, performance and partnership considerations. Replit announced a partnership with OpenAI in August.
These statements do not, however, establish a complete split. TechCrunch reports that a person close to the partnership, speaking anonymously, says it remains in place. The outlet also notes that a Replit post published in late September still presented Claude Fable 5 and Claude Fable 5.1 as models used in Replit Agent’s production environment.
Masad believes the Sol and Astra models deliver better results when building applications from scratch. He cites internal evaluations and a public benchmark. However, the article provides neither detailed results nor the methodology needed to examine this comparison: it remains an assessment attributed to the executive.
The commercial issue concerns the choice of model providers for a platform that allows applications to be developed and deployed from a browser. The available evidence leaves unresolved the distinction between a commercial preference, a technical change and the actual end of a partnership. TechCrunch says it asked Replit for clarification; its article reports no response resolving this contradiction.