Although Replit is calling the feature Free Mode, it still requires a paid subscription—$20 per month for the Core tier or $100 per month for Pro. But the idea is to inject more value into these subscriptions, and to remove the stress of burning through AI tokens. A user may, for example, chat, ideate, or run a simple task on Free Mode before tapping into the larger, limited budget of their subscription.
Free Mode is now the default for Core and Pro users, an approach Michele Catasta, the president and head of AI at Replit, calls “radical.”
He said the new service was made possible by OpenAI slashing costs on the Luna model by 80% on July 30.
“A lot of tasks don’t require the frontier-level intelligence,” Catasta says. “It just requires smaller models that are faster and more affordable.” He added that GPT-5.6 Luna is both “very powerful and reliable.”
When the Replit agent determines the task requires a different or more sophisticated model, it automatically re-routes the request “for the duration to get the task done, and then we fall back into Free Mode as soon as possible,” Catasta said.
A race to the bottom or more ROI?
Replit’s new feature is part of a wider trend within the AI industry of companies looking to lower the cost of AI products as enterprise customers are increasingly concerned about the return on investment from their AI spend, and as competition among American AI model builders heats up amid competition from cheap, Chinese AI models.
To that end, OpenAI slashed costs by a whopping 80% on its GPT-5.6 Luna model for developers using the API last month. The price cuts were driven by greater efficiencies in running the model, not an influx of new compute supply, Thibault Sottiaux, Head of Core Products and Platform at OpenAI, tells Fortune.
OpenAI is also working to add more value into ChatGPT. “Month after month we push on the amount of utility you can get for the same dollar amount,” Sottiaux said. “It’s all about using the right tool and the right model for the right outcome.”
OpenAI and Replit tease more joint launches to come
Although developers can typically access many models through Replit, GPT-5.6 Luna is the only model powering Free Mode.
“All I’m going to say is that we’re working more and more closely with OpenAI, and we’re really excited about it,” he said. “This is the first of many launches we’re going to be doing together.”
Replit and OpenAI have been working together for years, with OpenAI noting that Replit was an “early user of GPT-3,” which came out in in 2020. Replit was founded in 2016 by its CEO Amjad Masad. Catasta and Sottiaux said the companies share a vision to make powerful, affordable models accessible to as many people as possible. That includes empowering people without a technical background to vibe code.
OpenAI CEO Sam Altman imagines one day AI will be so ubiquitous it will be considered a public utility, metered to the public like water or electricity, Business Insider reported in March.
“If we can get to a world where anyone with access to the internet can build a product, build a startup, and just kind of get started and get going, I think we’re going to see another renaissance-level entrepreneurial boom like we’ve never seen before,” said Altman in statement concerning the Replit Free Mode announcement.
Read more Current price of oil as of August 19, 2026
About the Author
