Why it mattersA team that built on Replit's Claude integration cannot plan around what the models are if the company and its partner are now telling different stories.
A team building on Replit does not know this week which model is running underneath, because the company and its biggest partner are telling different stories.
TechCrunch reports that Replit CEO Amjad Masad said on stage at TechCrunch Disrupt that his company is no longer working with Anthropic. He said the decision came down to cost, partnership and performance, and added that ChatGPT is better now. Masad said Replit publishes a public benchmark and that OpenAI's Sol and Astra models score higher on building vibe-coded apps from scratch than Anthropic's Claude on it.
What is in dispute
TechCrunch quotes a person familiar with the arrangement, who asked not to be named, saying the Anthropic partnership is still in place. The outlet points at Replit's own blog from late September, which names Claude Fable 5 and Claude Fable 5.1 as production models in Replit Agent. Around the same time one of Replit's product employees posted on X promoting Opus 5.5. TechCrunch says it reached back out to Replit for clarification and will update the piece if it hears back.
That leaves two public statements from Replit people, both recent, that contradict each other. The CEO says Anthropic is out. The company's blog and a staff post treat Anthropic's models as current.
The context Masad gave
Masad also said enterprise has gone from about 2 percent of Replit's revenue last year to about 15 percent this year, and that more than 50 percent of revenue is now B2B. He did not share annual recurring revenue. On safety he said agents need a habitat with good orchestration, error recovery, sandboxing, networking and identity, and that this habitat is the thing to get right before deploying agents.
Replit announced an OpenAI partnership in August. Until this week there was no public word that the Anthropic relationship had ended. The breakup, if it is one, is also unusual on price: Replit has been one of the public reference customers for Claude in coding, and switching providers at this scale means re-benchmarking agent behaviour, re-tuning prompts, and accepting that users will notice differences in how the Agent writes.
For a team building on top of Replit, the practical read is to find out which model your Replit workloads are running on before this week's deploys, and ask the account team in writing. Treat vendor benchmark claims from either side as marketing until a third party runs the same test on your code. If your contract tied the price or the terms to a specific model family, the switch is a reason to re-read the contract.
Source
Replit is no longer working with Anthropic, CEO says by Lucas Ropek, TechCrunch.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
