The partnership claim is unsettled
Masad said Replit evaluates models with an internal benchmark and also publishes an open one. For building apps from prompts, he said, Sol and Astra consistently perform best. Replit offers a cloud development platform where users can build and run applications in a browser; its partnership with Anthropic had put Claude models on that platform. The collaboration was also reported to have helped Replit grow revenue significantly.
TechCrunch said it had asked Replit for clarification and would update its report if it received a response. The September post naming Claude Fable 5 and 5.1 does not by itself settle whether the partnership continued after that. But it makes the timing of Masad’s claim important.
Enterprise is becoming the bigger business
Masad also said enterprise customers are contributing a growing share of revenue. Replit has not disclosed annual recurring revenue, but the share from enterprise contracts rose from 2% last year to 15% this year. More than 50% of revenue now comes from B2B.
That shift gives the partnership question more weight. Replit is no longer describing itself only as a tool for individual developers: its business increasingly depends on serving organizations, where model choices and the infrastructure around them matter.
Agents need a safe operating environment
Masad compared AI agents’ working conditions with people’s: people need an office, food, air conditioning and a computer to do their jobs well. Agents, he said, need orchestration, recovery from errors, protection and security, isolation, networking, and identity and permission management. With that environment in place, AI can be deployed safely.
I think the most consequential gap in the announcement is not which model wins Replit’s benchmark, but what “no longer working with Anthropic” means in practice. The September production reference points to continued use at that time; the unnamed source says the collaboration continues. Until Replit clarifies the relationship, its model-selection story remains harder to verify than its growing enterprise numbers.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X