Replit CEO says company no longer working with Anthropic

Replit, the AI coding platform popular among startup founders, is no longer working with Anthropic, according to CEO and co-founder Amjad Masad. Masad made the announcement while onstage at TechCrunch Disrupt on Thursday, stating that the decision was driven by a combination of cost, partnership dynamics, and performance metrics. This marks a significant shift for Replit, which previously relied on Anthropic’s Claude models to power its cloud-based development environment, a relationship that was credited with helping the company grow its revenue significantly.

Masad clarified that Replit had announced a partnership with OpenAI in August, but there had been no public indication that the Anthropic relationship had ended. "I think ChatGPT is better now," Masad said, noting that the company uses internal benchmarks to assess performance. He added that Replit publishes a public benchmark and that their models, Sol and Astra, consistently perform better on building vibe-coded applications from scratch. "It was a decision based on both cost, partnership, and performance. And for all these reasons, we just felt like they were a better partner for now."

The details

Despite Masad’s onstage declaration, the situation appears more complex than a clean break. TechCrunch reports that insiders familiar with the partnership rejected the notion that it had ended. One person familiar with the matter, who asked not to be named, stated that the partnership was still in place. This contradiction is supported by recent public evidence from Replit’s own channels. A blog post from late September explicitly cites Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment. Around the same time, a Replit product employee posted on X about using Anthropic’s flagship model, Opus 5.5, and promoted its capabilities. TechCrunch reached out to Replit for clarification but had not received a response as of the publication of the report.

Context

Replit has positioned itself as a browser-based software development platform where users can develop and deploy apps directly from the interface. The company’s previous reliance on Anthropic’s models was a key part of its value proposition, offering developers access to high-quality AI coding assistance without managing local infrastructure. The shift to OpenAI, announced in August, suggests a strategic pivot toward what Masad perceives as a superior partner in terms of cost-efficiency and model performance for specific use cases, particularly "vibe-coded" applications. This move aligns with a broader trend among AI coding tools to diversify or switch model providers based on benchmark performance and pricing structures, rather than maintaining exclusive relationships with a single lab.

What it means for developers

For developers using Replit, the immediate impact may be subtle, as the platform likely maintains a multi-model approach. However, the public dispute over the status of the Anthropic partnership raises questions about the stability of Replit’s backend infrastructure. Users who specifically relied on Claude models for their workflows should monitor Replit’s documentation and blog posts for updates on model availability. The mention of "Sol and Astra" performing better on specific benchmarks suggests that Replit is actively curating its model lineup based on task-specific performance, which could mean that certain types of coding tasks will be routed to different models depending on the provider. Developers should be aware that the "best" model for a given task may change as Replit adjusts its partnerships and internal benchmarks.

What to watch

• Clarification from Replit: Whether Replit issues an official statement confirming or denying the end of the Anthropic partnership.
• Model Availability: Updates to Replit’s documentation regarding which models are available in the production environment, specifically if Claude models are removed or deprecated.
• Benchmark Updates: Changes to Replit’s public benchmarks that might explain the performance gap between OpenAI and Anthropic models for their specific use cases.
• Enterprise Adoption: How this shift affects Replit’s enterprise contracts, given that Masad noted enterprise revenue has grown from 2% to 15% of total revenue this year.