Replit makes OpenAI its default over Anthropic, CEO says

Replit Switches Default AI Model to OpenAI

Replit, the cloud‑based development environment that has become a staple for many startup founders and developers, announced at TechCrunch Disrupt that it is now using OpenAI’s ChatGPT as the default AI model for its platform. The shift, explained by CEO and co‑founder Amjad Masad, reflects a combination of cost considerations, performance benchmarks, and strategic partnership factors. While OpenAI will be the default, Replit remains model‑agnostic and continues to offer Anthropic’s Claude models as an option for users who prefer them.

---

Why OpenAI Became the Default

Masad highlighted three primary reasons for the change:

1. Cost Efficiency – Running large language models at scale can be expensive. OpenAI’s pricing structure, combined with the volume of usage on Replit, offers a more sustainable financial model for the company.
2. Performance Benchmarks – Replit runs internal and public benchmarks that compare model capabilities on tasks such as building "vibe‑coded" applications from scratch. According to Masad, OpenAI’s models consistently outperformed Anthropic’s on these metrics.
3. Partnership Strength – The recent partnership announced in August has deepened collaboration between the two companies, giving Replit confidence in OpenAI’s roadmap and support.

Masad emphasized that the decision was not a wholesale abandonment of Anthropic. "Replit remains model agnostic," he said, noting that users can still select Claude or other supported models if they wish.

---

Anthropic Still on the Platform – But Not the Default

Despite the default switch, Replit’s platform continues to host Anthropic models. The company’s documentation and recent blog posts from September still reference Claude Fable 5, Claude Fable 5.1, and the Opus 5.5 model. An employee’s social‑media post also highlighted Anthropic’s capabilities, indicating that the integration remains functional.

The practical impact of the default change is significant: most users tend to stick with the pre‑selected model, meaning Replit’s spend on Anthropic is likely to drop. However, the ability to manually select Anthropic models preserves flexibility for developers who have specific preferences or requirements.

---

Growth Signals and Enterprise Focus

Masad used the Disrupt stage to share broader company metrics. While exact ARR figures were not disclosed, he noted a rapid shift toward enterprise revenue:

• Enterprise accounted for roughly 2 % of revenue last year.
• That share has risen to about 15 % this year, with over half of Replit’s total revenue now coming from B2B customers.
• Looking ahead, Masad expects more than 50 % of revenue in 2027 to be driven by enterprise and government contracts.

These numbers suggest that Replit is positioning itself as a serious AI‑enabled development platform for larger organizations, not just individual developers or hobbyists.

---

Safety, Orchestration, and the "Habitat" for AI Agents

Beyond the business rationale, Masad stressed the importance of a robust safety and operational framework for AI agents. He likened the requirements for AI to a physical workplace: agents need a well‑orchestrated environment that includes error recovery, sandboxing, secure networking, and clear identity/authorization controls. This "habitat" is essential for deploying AI safely at scale, especially as Replit expands its enterprise footprint.

---

What This Means for Developers

• Default Experience: New users will see ChatGPT responses by default when they invoke Replit’s AI features.
• Model Choice: Power users can still switch to Anthropic’s Claude or other supported models via the platform’s settings.
• Cost Implications: Projects that rely heavily on AI may see lower per‑token costs, potentially enabling more extensive experimentation.
• Enterprise Tools: Companies adopting Replit for internal tooling can expect tighter integrations with OpenAI, alongside Replit’s focus on security and compliance.

---

Looking Forward

Replit’s decision underscores a broader trend in the AI tooling space: platforms are gravitating toward the providers that combine performance, cost‑effectiveness, and strong partnership ecosystems. While OpenAI currently holds the default spot on Replit, the company’s commitment to model agnosticism ensures that developers retain the freedom to choose the best tool for their specific needs.

As Replit continues to grow its enterprise business and refine its AI safety infrastructure, the platform is poised to become a central hub for building, testing, and deploying AI‑augmented applications—whether they run on OpenAI, Anthropic, or future model providers.