Replit CEO Amjad Masad says the coding platform is no longer working with Anthropic, pointing to OpenAI’s models as cheaper and better for its needs. The claim about the Replit Anthropic partnership was quickly disputed by a person familiar with the deal, who said it is still in place. Recent Replit posts also show Claude models running inside its product only weeks ago.
What Masad Said About the Replit Anthropic Partnership
Masad said the move came down to a mix of cost, partnership terms and performance. Replit announced a partnership with OpenAI in August, but at the time nothing suggested it was leaving Anthropic behind. He told a reporter he thinks ChatGPT is “better now.”
According to Masad, Replit’s internal and public benchmarks show OpenAI’s Sol and Astra models doing better at building vibe-coded apps from scratch. He framed OpenAI as the better partner “for now,” which leaves the door open for future changes.
Why the Anthropic Split Is Being Questioned
An unnamed person familiar with the relationship rejected Masad’s account and said the partnership remains active, according to the original report. There is also public evidence that Replit leaned on Claude very recently.
- A late-September Replit blog post listed Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production setup.
- Around the same time, a Replit product employee posted on X praising Anthropic’s flagship Opus 5.5 model.
- Separate coverage of Masad’s remarks says Claude has been central to Replit Agent since 2024 and that the integration remains active.
Replit had not clarified the conflicting accounts at the time of the report. So for now, it is not clear whether Claude has been fully removed or simply moved out of the spotlight.
Replit’s Multi-Model Strategy Beyond Anthropic
Whatever the final answer, Masad was clear that Replit does not want to depend on any single AI lab. That matters because Anthropic and OpenAI both build their own coding tools that compete for the same users. Keeping the option to swap or mix models is Replit’s hedge against that risk.
| Detail | What has been reported |
|---|---|
| Revenue in 2024 | About $2.8 million |
| Projected 2026 run rate | Approaching $1 billion |
| Valuation | Rose from $3 billion to $9 billion this year after a $250 million round |
| Enterprise share of revenue | About 15%, up from roughly 2% last year (per Masad) |
| B2B share of revenue | More than 50% (per Masad) |
The “Self-Driving Company” Pitch
Masad also pitched what he calls a self-driving company. At Replit, AI agents do the first pass on support tickets, digging through code and issue trackers before a person steps in. He argues agents need a safe “habitat” with sandboxing, error recovery and proper identity controls before they can be trusted at scale.
What This Means for You
If you build apps on Replit, you probably will not notice a sudden change. The platform’s whole pitch is model flexibility, so the AI behind the scenes can shift without you choosing it. A few practical points to keep in mind:
- Results from the same prompt may change as Replit swaps or blends models.
- If you prefer Claude for coding, tools like Claude AI and Claude Code are still available directly.
- Comparing platforms is easier with our guide to Cursor vs Replit for new coders.
- Watch for an official statement from Replit or Anthropic before assuming the split is final.
The dispute also shows how much pressure AI app makers face on model costs. Accenture recently capped Claude spending per employee, a sign that price now shapes which models companies pick. For more updates, see our latest AI news.
Sources
This article draws on reporting from TechCrunch on Masad’s comments and the pushback, and from Crypto Briefing on Replit’s growth figures and multi-model strategy.
Conclusion
Replit’s CEO says the company has stopped working with Anthropic and now favors OpenAI on cost, partnership and performance. A person familiar with the deal disputes that, and Replit’s own recent posts show Claude models in production just weeks ago. What is certain is that Replit wants freedom to switch between AI providers as labs become rivals. With revenue climbing toward a $1 billion run rate, its model choices now carry real weight. Until Replit or Anthropic clarifies, treat the breakup as Masad’s claim rather than settled fact.
