Replit Switches to OpenAI as Default Model Provider

Replit, the AI coding platform favored by startup founders and developers, has shifted OpenAI to its default model provider, replacing Anthropic. The change signals how developer-focused platforms are now making strategic decisions between frontier models for AI-assisted coding.

CEO and co-founder Amjad Masad announced the decision while speaking at TechCrunch Disrupt on October 17. He cited three factors: performance, cost, and partnership quality. Masad explained that Replit uses internal benchmarks and public benchmarks to compare models, and that OpenAI’s offerings consistently outperform alternatives when building applications from scratch.

Masad told the audience that “ChatGPT is better now” and described the decision as grounded in data. The company has published benchmarks to support this assessment.

This represents a substantial shift for Replit. Until recently, the platform ran on Anthropic’s Claude models. That partnership was significant. Masad disclosed that Replit had been a top-five customer of Anthropic, and the relationship helped drive meaningful revenue growth for the coding platform.

The timeline adds context. Replit announced a partnership with OpenAI in August, but there was no public indication at that time that the Anthropic relationship would change. The switch from Anthropic to OpenAI as the default became official at Disrupt in mid-October.

The practical impact centers on default selection. While Masad clarified after the event that users can still access Anthropic’s Claude models if they choose, most customers will stick with the default option. This means Replit’s spending with Anthropic will likely drop substantially even though the models remain available on the platform.

Masad’s public statements onstage drew pushback from people close to the Anthropic partnership. An unnamed source told TechCrunch that the partnership was still active, contradicting the impression Masad gave during his presentation. Evidence supports this claim. Replit’s website, blog posts, and social media showed recent promotions of Anthropic products. A late-September blog post cited Anthropic’s Claude models as running in Replit Agent’s production environment. Around the same time, Replit staff were publicly promoting Anthropic’s capabilities on social media.

The contradiction highlights that the announcement lacked clarity about what changing the default relationship actually means. Masad later clarified the company’s position: “Replit remains model agnostic. Our users can choose Anthropic’s Claude or other supported models. OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.” This reframing suggests the shift is about default selection and partnership priority, not a complete exit.

The move reflects a broader pattern in AI tooling. Platform builders making default model choices now evaluate frontier models against multiple criteria: benchmarking, costs, and the depth of collaboration with each vendor. For Replit, the calculation weighed performance against economics.

What changed

Replit changed its default model from Anthropic’s Claude to OpenAI’s ChatGPT, announced at TechCrunch Disrupt on October 17. Users can still choose Claude or other models, but OpenAI is now the primary option. Replit was previously a top-five customer of Anthropic before this shift.

Why it matters

Replit’s choice will influence how many developers first experience different AI models. Developers using Replit Agent without actively selecting an alternative will now interact with OpenAI’s offerings. This could shift the feedback loop toward OpenAI’s strengths and limitations.

The shift also carries financial implications for Anthropic. Losing a top-five customer represents material revenue impact, particularly if the default change reduces usage even among customers who could technically choose Claude. For OpenAI, it validates an embedding strategy focused on integrating its models into developer tools.

Masad also used the Disrupt appearance to discuss Replit’s business trajectory. Enterprise revenue grew from 2% last year to 15% of total revenue this year. More than 50% of revenue now comes from B2B contracts. Masad told TechCrunch he expects enterprise and government customers to account for more than half of all revenue in 2027. This suggests Replit is shifting toward becoming an enterprise development platform.

He also addressed AI safety, arguing that autonomous agents require proper infrastructure to operate safely. “Agents need a habitat,” Masad said, describing the requirements as good orchestration, error recovery, safety, security, sandboxing, networking, and identity management. This framing aligns with industry-wide concerns about deploying autonomous systems reliably.

What to test

The decision raises several questions worth monitoring. Does OpenAI’s default status translate into higher customer satisfaction on Replit compared to the Anthropic era? Do developers who stay on the default report different coding experiences than those actively selecting Claude?

Another question: Will other developer platforms make similar switches? If Replit’s decision proves durable and leads to strong platform metrics, competitors may follow. Conversely, if Claude continues to outperform OpenAI on specific coding tasks, platforms may revise their defaults. Model selection for specialized workflows like coding is empirical and can shift if performance or costs change materially.

The conclusion

Replit’s shift from Anthropic to OpenAI as the default model reflects how developer platforms are now making strategic decisions based on benchmarking, cost, and partnership depth. While Anthropic models remain accessible, the default change will likely reduce Replit’s overall Anthropic usage since most users stick with the default option. The decision matters because platform defaults shape which models developers encounter first and how they form initial impressions. For Anthropic, losing a significant customer is material. For OpenAI, embedding its models into tools like Replit validates a strategy focused on integration rather than standalone model competition.

Related reading

• JetBrains Mellum2.1: 12B MoE Open Model for Coding Agents
• RTX Spark Windows Laptops Now Available Starting at $2,399
• Mistral Large 4 vs Meta Muse Wonder: Open vs Closed