Replit Drops Anthropic for OpenAI, Says CEO

Replit announced it had ended its partnership with Anthropic, the creator of Claude, during a speech at TechCrunch Disrupt. The move follows a recent August alliance with OpenAI and follows cost, performance and partnership considerations, according to CEO Amjad Masad. Masad also highlighted a jump in enterprise revenue share from 2% last year to 15% this year, underscoring the financial stakes of the decision.

Key points

• Replit ends Anthropic partnership after partnership with OpenAI announced in August.
• CEO cites cost, performance and partnership as reasons for the switch.
• Enterprise revenue share climbed from 2% to 15% year‑over‑year.
• Replit’s platform uses AI agents; safety and sandboxing are key to deployment.
• Evidence of Anthropic usage persisted into September, raising questions about the departure.

What happened

During the TechCrunch Disrupt keynote, Masad said the company’s internal benchmark showed OpenAI’s Sol and Astra outperforming Claude in building vibe‑coded applications from scratch. The report quotes CEO Amjad Masad saying "I think ChatGPT is better now," and that the decision was based on cost, partnership and performance. Replit had announced its OpenAI partnership in August, but no prior public signal suggested a break with Anthropic.

Replit’s platform lets developers build and deploy apps directly from a browser. Under the Anthropic partnership, the company ran Claude models in its Replit Agent environment, a collaboration that reportedly helped grow revenue. However, insiders familiar with the partnership said it was still active, and a late‑September blog post cited Claude Fable 5 and 5.1 running in production. A Replit employee also promoted Anthropic’s Opus 5.5 on X.

The backstory

Replit has long positioned itself as a low‑friction, cloud‑based IDE, attracting startup founders and developers with its instant‑on, browser‑only workflow. The company’s move toward AI integration began with a 2025 partnership with Anthropic, leveraging Claude to accelerate code generation and debugging. The partnership was marketed as a revenue driver, and Replit’s leadership highlighted the ability to scale AI workloads in a sandboxed, secure environment.

In August 2026, Replit announced a new alliance with OpenAI, announcing the integration of GPT‑4 and ChatGPT into its platform. The announcement coincided with a surge in enterprise contracts, which Masad said now account for 15% of revenue compared to 2% the previous year. The shift to OpenAI was framed as a cost‑effective, high‑performance alternative, but the public narrative did not immediately clarify the status of the Anthropic relationship.

Why the switch matters

Cost was a primary driver. Anthropic’s pricing model for Claude, especially at higher token volumes, was reportedly higher than OpenAI’s tiered structure. Replit’s internal benchmark showed OpenAI’s models delivering faster inference and higher accuracy on the platform’s core use cases. The CEO’s emphasis on a “public benchmark” suggests Replit is actively measuring AI performance against business metrics.

The partnership shift also signals a broader industry trend. Enterprises increasingly demand turnkey AI solutions that integrate seamlessly with existing cloud stacks and comply with stringent security guidelines. Replit’s focus on sandboxing, error recovery, identity, and authorization reflects a growing emphasis on safe AI deployment—a factor that could sway enterprise buyers.

Who wins and who loses

Replit gains a partner with a larger ecosystem, potentially lowering costs and expanding feature parity with competitors. The company’s move to OpenAI may accelerate adoption among developers who already use ChatGPT, strengthening its value proposition. Anthropic, however, loses a high‑profile platform that had helped validate Claude’s commercial viability.

Enterprises that have committed to Replit’s Anthropic‑based workflows may face migration work, but the shift could also reduce their AI spend. Developers who rely on Replit’s AI agents will need to adapt to new model interfaces, but might benefit from improved performance and broader support.

What builders and businesses should do

• Evaluate your AI model costs across providers; benchmark performance on your specific workloads.
• Plan for migration paths if your platform partners shift; maintain modular integration layers.
• Monitor enterprise‑grade AI safety features—sandboxing, error recovery, and identity controls.
• Engage with platform vendors to understand long‑term partnership commitments.
• Keep an eye on revenue mix reports; a 15% enterprise share may indicate a shift in strategic focus.

The bottom line

Replit’s pivot to OpenAI highlights the importance of cost and performance in AI platform partnerships. The next few months will reveal whether the switch delivers the promised savings and speed gains, and whether Anthropic can regain footing in the developer tools market. Businesses relying on Replit should track model performance metrics and negotiate flexible contract terms to mitigate future partner churn.