Replit Integrates OpenAI's Luna Model for New 'Free Mode'

Replit, a company focused on collaborative coding environments, has launched a new feature called 'Free Mode,' which is powered by OpenAI's GPT-5.6 Luna model. This announcement, shared exclusively with Fortune, signifies an expansion of the partnership between Replit and OpenAI, with plans for future product development.

While named 'Free Mode,' the feature is integrated into Replit's existing paid subscription tiers: the Core tier at $20 per month and the Pro tier at $100 per month. The objective of Free Mode is to provide increased value within these subscriptions and alleviate user concerns about depleting their AI token allowances. Users can utilize Free Mode for tasks such as chatting, brainstorming ideas, or executing simple operations before opting to use the more substantial, but limited, resources of their paid subscription. Michele Catasta, president and head of AI at Replit, described this approach as "radical," noting that Free Mode is now the default experience for both Core and Pro subscribers.

The implementation of Free Mode was made possible by a significant cost reduction of 80% on OpenAI's GPT-5.6 Luna model, which occurred on July 30. Catasta explained that many computational tasks do not necessitate the most advanced, frontier-level AI capabilities. Instead, they can be efficiently handled by smaller, faster, and more affordable models like GPT-5.6 Luna, which he characterized as both "very powerful and reliable." In instances where the Replit agent identifies a task requiring a more sophisticated model, it automatically reroutes the request for the duration of that task. Once completed, the system reverts to Free Mode to conserve resources.

Replit's introduction of Free Mode aligns with a broader trend in the artificial intelligence industry, where companies are actively seeking to reduce the operational costs of their AI products. This strategic shift is driven by increasing concerns among enterprise clients regarding the return on investment (ROI) for their AI expenditures. The competitive landscape among AI model developers in the United States is intensifying, further influenced by the availability of low-cost AI models originating from China. In response to these market dynamics, OpenAI implemented an 80% price reduction on its GPT-5.6 Luna model for developers utilizing its API. According to Thibaut, these cost cuts were achieved through enhanced operational efficiencies in running the model, rather than an increase in available computing power.