Replit's shift from Anthropic to OpenAI sparks debate

Replit, an AI coding platform favored by developers and startups, has reportedly ended its partnership with Anthropic, according to CEO Amjad Masad, who made the announcement at TechCrunch Disrupt. Masad cited cost, partnership, and performance as key factors, noting that ChatGPT is now superior and that Anthropic's models, Sol and Astra, underperformed in benchmark tests. Replit previously partnered with Anthropic to integrate Claude models into its platform, which contributed to significant revenue growth. However, insiders dispute the claim, stating the partnership remains active. Evidence on Replit's website and social media suggests continued use of Anthropic models as recently as late September. Masad also highlighted the growing importance of enterprise revenue, which now accounts for 15% of Replit's income, up from 2% last year. He emphasized the need for robust AI safety measures, comparing them to the environment required for human productivity. The article was updated to include additional context on Replit's use of Anthropic models.