Replit CEO claims partnership with Anthropic ended, citing costs
Replit leadership stated the AI coding platform stopped using Anthropic models, though insiders and recent evidence suggest the partnership remains active.
Key points
- Replit CEO Amjad Masad stated the company is no longer working with Anthropic during a speech at TechCrunch Disrupt 2026.
- Masad cited cost, partnership dynamics, and performance metrics as reasons for the shift toward OpenAI.
- Insiders familiar with the deal told TechCrunch the partnership is still in place, citing September blog posts.
Replit CEO and co-founder Amjad Masad stated that his company is no longer working with Anthropic, according to remarks made on stage at TechCrunch Disrupt 2026. Masad announced the change in partnership while discussing the platform's strategic direction and recent growth metrics with the audience.
Masad attributed the decision to end the collaboration with Anthropic to a combination of cost, partnership terms, and performance issues. He noted that Replit uses internal benchmarks to assess model capabilities, claiming that their own Sol and Astra models consistently outperform others in building applications from scratch.
Conflicting Reports on Partnership Status
Despite Masad’s public statements, sources familiar with the arrangement disputed the claim that the partnership has ended. According to TechCrunch, one person familiar with the situation said the partnership between Replit and Anthropic remains in place. This insider account directly contradicts the CEO’s assertion that the companies are no longer working together.
Evidence suggests Replit was actively using Anthropic technology as recently as September. A blog post from that month cited Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment. TechCrunch stated it reached out to Replit for clarification and would update the story if it received a response.
Strategic Shift and Business Growth
Masad highlighted that Replit announced a partnership with OpenAI in August, though he did not explicitly link the end of the Anthropic deal to the start of the OpenAI one. He remarked that ChatGPT is better now, suggesting a preference for OpenAI’s offerings based on current performance metrics. This shift aligns with the company’s broader strategy to optimize its AI infrastructure.
The CEO also shared insights into Replit’s financial composition, noting a significant increase in enterprise revenue. Masad stated that enterprise contracts accounted for only two percent of revenue last year but now represent fifteen percent. He added that more than fifty percent of the company’s total revenue now comes from business-to-business contracts.
What should you do?
Security professionals and developers using Replit should monitor official channels for confirmation regarding the underlying AI models powering their workflows. Users relying on specific model behaviors for security or compliance checks should verify which provider is currently active in their environment.
What should you do?
What to do and how to stay safe: Replit
- Verify which AI models are currently powering your Replit workflows by checking official documentation or support channels.
- Monitor for official statements from Replit or Anthropic to confirm the status of their partnership.
- Audit your application dependencies to ensure they remain compatible with potential changes in underlying AI providers.
- Review your service level agreements with Replit to understand how changes in AI partners might affect your deployment.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Did Replit officially confirm it ended its partnership with Anthropic?
No, while CEO Amjad Masad stated the partnership ended, insiders told TechCrunch it is still in place, and TechCrunch is seeking clarification.
What reasons did Replit give for stopping work with Anthropic?
Masad cited cost, partnership dynamics, and performance metrics, stating that their internal models and OpenAI’s ChatGPT performed better.
Was Replit using Anthropic models recently?
Yes, a September blog post cited Anthropic’s Claude Fable 5 and 5.1 as running in Replit Agent’s production environment.
