Replit has made OpenAI its preferred AI model partner, but Anthropic's Claude models remain available on the AI-powered software development platform. The distinction emerged after Replit CEO and co-founder Amjad Masad said at TechCrunch Disrupt on October 15, 2026, that his company was no longer working with Anthropic.
Masad subsequently clarified that Replit continues to support multiple AI providers and that customers can still choose Claude. His comments have nevertheless raised questions about the status of Replit's commercial relationship with Anthropic, particularly after a person familiar with the arrangement disputed the suggestion that the partnership had ended.
Why Replit Is Favoring OpenAI
Masad attributed the preference for OpenAI to a combination of performance, economics, and the companies' working relationship. Speaking at TechCrunch Disrupt, he said Replit evaluates models using internal testing as well as publicly reported benchmarks to help determine which systems are best suited to building applications.
In subsequent comments reported by TechCrunch, Masad described OpenAI as Replit's preferred partner for powering Replit Agent while emphasizing that the platform remains model-agnostic.
The shift builds on an existing relationship. On August 18, 2026, Replit announced an expanded OpenAI-powered experience alongside changes to its AI development tools. That release introduced Free Mode powered by OpenAI's GPT-5.6 Luna and described a broader effort to make AI-assisted software creation more accessible and cost-efficient.
Replit has also published technical research examining how its Agent architecture coordinates AI models, tools, and subtasks. Its engineering analysis of agent performance discusses benchmark results and production configurations involving models from different providers.
Those findings provide context for Replit's technical decisions, but they should not be interpreted as independent proof that OpenAI models outperform Anthropic's Claude models across every programming task. AI coding performance can depend on the specific model, task, execution environment, evaluation method, and software tools surrounding the model.
Likewise, although Masad cited economics as a factor, Replit has not publicly established a verified dollar figure for the savings attributable to its preferred-provider change.
Can Developers Still Use Claude on Replit?
Yes. Replit's official documentation continues to list supported Anthropic models alongside OpenAI models, although access depends on the user's plan, selected Agent mode, and account availability.
According to Replit's model-selector documentation, the platform supports two approaches: intelligent model routing, which automatically chooses models according to task requirements and cost considerations, and manual selection from available models.
Free Mode uses intelligent routing and does not provide a manual model selector. Core and Pro subscribers can choose available models in Power Mode or Max Mode, subject to the features enabled for their accounts.
The documented selection includes Claude Sonnet models in Power Mode and additional Claude models, including Claude Opus and Claude Fable variants, in Max Mode. Replit also lists OpenAI models in these modes. Actual availability can vary by rollout, organization settings, authorization, and subscription restrictions.
This means developers should not interpret OpenAI's preferred status as a platform-wide ban on Claude. However, the existence of a manual model selector also does not establish that every Replit feature or automatically routed task offers the same level of model choice.
For teams that depend on a particular model, the most useful immediate step is to review the available models and controls within their own Replit workspace. The company's published documentation does not establish that this partnership development requires existing users to migrate projects or abandon Claude-based workflows.
What Is Known About the Anthropic Partnership?
The distinction between model availability and the underlying commercial partnership is central to the story.
Masad's initial remarks suggested Replit had stopped working with Anthropic. However, his subsequent clarification established that customers could continue selecting supported Claude models and that OpenAI had become the preferred partner rather than the platform's only available provider.
TechCrunch also reported that a person familiar with the Replit-Anthropic relationship, speaking anonymously, disputed the suggestion that the partnership had ended and maintained that it remained in place.
These accounts do not conclusively resolve the contractual status of the relationship. Continued model access demonstrates that Anthropic technology remains available through Replit, but it does not independently establish whether a broader commercial agreement has been renewed, modified, reduced, or terminated.
The companies also have a substantial operating history. Before Replit expanded its OpenAI partnership, Anthropic's Claude models played an important role in powering the coding platform. Masad said Replit had previously ranked among Anthropic's five largest customers, although that ranking has not been independently verified.
Replit's technical publications also identify Anthropic models used in production during August and September 2026. That recent usage makes the timing and scope of any commercial change particularly relevant, but it does not prove the present contractual position.
Without a definitive public clarification establishing the agreement's status, describing the development as a complete technological separation would overstate the available evidence.
What the Shift Means for AI Coding Platforms
Replit's decision illustrates how AI development platforms are becoming an important competitive channel for companies building large language models.
Rather than requiring developers to work directly with a model provider, services such as Replit integrate AI into a broader environment for generating code, modifying applications, testing software, and deploying projects. The platform determines how models are presented, routed, and combined with development tools.
That makes default-model selection commercially significant. When a platform favors one provider for routine workloads, the decision can influence how frequently that provider's models are used, even if customers retain the option to choose alternatives. The precise financial consequences for OpenAI and Anthropic, however, have not been publicly established in this case.
The development also reflects wider competition in AI-assisted programming. Companies including Google have expanded their offerings in this area, with tools designed to make software development accessible to more users. Earlier developments involving Google's AI coding tools illustrate how model capabilities and developer access have become competing priorities across the industry.
Replit's enterprise ambitions add another dimension. At TechCrunch Disrupt, Masad said enterprise contracts accounted for approximately 15% of company revenue in 2026, compared with approximately 2% in 2025, while business-to-business customers represented more than half of revenue. Those figures were reported by the executive rather than established through publicly audited financial statements.
For enterprise customers, a development platform's choice of AI providers may affect operational considerations such as model availability, administrative controls, consistency, and cost management. Replit's multi-model approach potentially offers flexibility, although specific capabilities remain dependent on product modes and account settings.
For now, the confirmed practical distinction is straightforward: OpenAI has become Replit's preferred partner, while Claude remains among the supported model options. Whether the commercial relationship between Replit and Anthropic has formally ended remains unresolved. Further statements from the companies, together with changes to Replit's published model availability, will determine whether this becomes a deeper separation or primarily a change in platform preference.
0 Comments