Replit makes OpenAI its default over Anthropic, CEO says

Replit has shifted its default model provider to OpenAI, moving away from its previous primary reliance on Anthropic for powering its browser-based software development platform and autonomous coding agent. Chief executive Amjad Masad initially stated on stage at TechCrunch Disrupt that the company was no longer working with Anthropic, citing performance benchmarks, operational costs, and the terms of its commercial agreements. As detailed in TechCrunch's report, Masad later clarified over the weekend that Replit remains model-agnostic, keeping Anthropic models accessible to users while establishing OpenAI as the preferred out-of-the-box engine.

This transition marks an operational reorganization for the cloud environment used by software engineers and startup teams to build and deploy applications directly from browsers. The development affects enterprise customers, individual developers writing code with Replit Agent, and model providers competing for high-volume developer infrastructure. An analysis of the company's internal benchmarks, shifting enterprise revenue mix, and platform configuration demonstrates how default tool settings alter commercial spend across foundation model vendors.

Replit makes OpenAI its default over: what actually changed

The primary operational change centers on the default model assignment in Replit Agent, which now routes user tasks through OpenAI rather than Anthropic. Although Masad initially claimed during a Disrupt appearance that Replit had severed ties with Anthropic, he later stated that the platform continues to support Claude models as optional selections. Insiders familiar with the commercial partnership between Replit and Anthropic rejected the assertion that collaboration had terminated, confirming that their formal business relationship remains active despite the change in default platform routing.

The downgrade from default engine to secondary option carries substantial financial consequences because the majority of Replit customers retain system defaults rather than manually changing backend settings. Replit previously ranked as one of Anthropic's top-five customers, having leaned on Claude architectures during a period of rapid business expansion. While promotional material and late-September blog entries highlighted Claude Fable 5, Claude Fable 5.1, and Claude Opus 5.5 in production workflows, OpenAI's ChatGPT technology now handles initial code-generation requests.

Replit makes OpenAI its default over: how it works

Replit operates an integrated browser development environment where software engineers write, compile, sandbox, and host web applications. Within this setup, Replit Agent coordinates multi-step development tasks, executing file edits, testing scripts, and debugging runtime errors inside dedicated cloud workspaces. To evaluate model capabilities for these workflows, Replit maintains both internal and public benchmarks designed around vibe-coded applications, scoring how cleanly different systems write operational software codebases from scratch without manual intervention.

According to Masad, OpenAI engines Sol and Astra demonstrated superior code execution and architectural resilience during these comparative benchmark evaluations. When evaluating runtime requirements for autonomous coding, Replit assesses whether underlying models can function within an orchestrated software habitat that demands automated error recovery, sandboxed network isolation, and granular identity controls. OpenAI emerged as the preferred platform engine because its operational economics and benchmark scores matched Replit's infrastructure requirements more closely than competing alternatives.

Replit makes OpenAI its default over: why it matters now

The platform shift highlights how model defaults dictate usage volumes in cloud development environments where software creators build entire applications through natural-language prompts. Prior to finalizing an August partnership with OpenAI, Replit drove a significant share of its computational traffic to Anthropic, helping generate substantial commercial revenue for the AI research lab. By substituting OpenAI into the default position, Replit substantially reduces its ongoing compute billing with Anthropic while channeling active software engineering pipelines to OpenAI.

The timing aligns with a shift in Replit's corporate financial composition away from individual hobbyists toward enterprise and business clients. Enterprise contracts accounted for roughly 2 percent of company revenue in 2025 before expanding to approximately 15 percent of revenue in 2026. With business-to-business agreements now generating more than 50 percent of overall turnover, Replit's procurement choices reflect enterprise priorities around tooling cost, system orchestration, and scalable cloud execution.

Replit makes OpenAI its default over: who is affected

Software engineers, startup founders, and technical teams relying on Replit Agent experience an altered baseline environment where OpenAI models execute commands by default. Developers who prefer Anthropic systems must now intentionally open settings menus to designate Claude Fable 5, Claude Fable 5.1, or Opus 5.5 for their production repositories. Teams building browser-deployed applications will see autonomous code generation steered by OpenAI's logic unless their organizational account profiles explicitly mandate alternate foundation models.

The vendor ecosystem experiences an immediate commercial realignment, given Replit's recent status as a top-five Anthropic enterprise buyer. Anthropic faces reduced API demand across Replit's browser-based IDE network as usage defaults swing toward OpenAI. Concurrently, enterprise administrators evaluating Replit's cloud infrastructure must account for OpenAI powering default automated code orchestration, sandboxed environment debugging, and identity management flows.

Replit makes OpenAI its default over: what to watch

Industry observers are tracking whether Replit achieves its long-term financial targets as corporate customer adoption accelerates. Masad projected that enterprise contracts and government accounts will collectively contribute more than half of the company's total revenue by 2027. Meeting that forecast will depend on whether business software buyers accept Replit's orchestrated agent habitats and whether the platform's security sandboxing meets enterprise compliance standards for cloud deployment.

Another focal point is the stability of Replit's multi-vendor support commitments over subsequent product cycles. While company leadership maintains that the environment will stay model-agnostic, Replit's technical benchmark publications will show how Sol, Astra, and newer Claude releases trade performance leads in autonomous software generation. Monitoring default retention rates will indicate whether developer teams remain content with preselected OpenAI models or continue manually selecting Anthropic options for specialized coding tasks.