Replit ends Anthropic partnership, cites cost and performance in shift to OpenAI

Replit’s CEO and co-founder Amjad Masad said the company is no longer working with Anthropic, attributing the decision to cost, performance, and partnership considerations. Speaking onstage at TechCrunch Disrupt on October 16, 2026, Masad said Replit uses internal benchmarks to assess model performance and that OpenAI’s models—including Sol and Astra—consistently performed better for building applications from scratch.

Masad added that the decision was based on "both cost, partnership, and performance," and that Replit felt OpenAI was "a better partner for now." Replit’s platform previously relied on Anthropic’s Claude models under a prior partnership, which was said to have contributed to significant revenue growth.

The announcement conflicts with evidence suggesting Replit continued using Anthropic’s models as recently as late September 2026. A Replit blog post cited Anthropic’s Claude Fable 5 and Claude Fable 5.1 as running in production within Replit Agent, and a product employee posted on X about using Anthropic’s Opus 5.5 model.

TechCrunch sought clarification from Replit but had not received a response at the time of publication. Masad also discussed Replit’s revenue mix, noting that enterprise contracts now account for 15% of revenue this year, up from 2% last year, with over 50% of total revenue coming from B2B sources.