Replit Users Face Higher Costs and Frustrations with New AI Updates
453If you’re using Replit’s vibe coding tools, you might have noticed your bills are going up. Many users have taken to Reddit this week to complain about increased costs, especially after Replit released its Agent 3 update earlier this month. Developers relying on AI helpers like Cursor, Kiro, and Claude Code are seeing their credits drain faster than before, sparking frustration across the community.
Why Are Costs Rising with Agent 3?
Replit says it’s working on fixing the issues, but many users are feeling the pinch. The main reason for the cost hikes is the new Agent 3, which is now more autonomous—about ten times more, according to Replit. This means it can generate other agents and automate tasks more aggressively. As a result, it often starts sub-agents that refactor code, even for small edits. This leads to higher compute usage and bigger bills.
Some users have reported burning through a third of their monthly credits in just one night. They’re also annoyed because the upgraded agent sometimes makes changes they didn’t request or want, which adds to their costs without adding value. Jason Andersen, a tech analyst, explained that refactoring code costs more than creating it from scratch. So, when the tool reworks working code, it’s not just frustrating—it’s expensive.
Changes in Pricing and User Control
Another factor behind the rising costs is Replit’s new effort-based pricing model introduced in July. Instead of counting checkpoints based on user requests, the system now charges based on how much compute power is used. Previously, a single request could produce multiple checkpoints, but now each request results in just one checkpoint, which might consume more resources. This shift means that even simple tasks can now cost more if they use a lot of compute.
Developers are also annoyed because Replit doesn’t let users pick an earlier version of the agent. That could help avoid some of the issues caused by Agent 3. Paul Chada, co-founder of DoozerAI, suggests Replit should consider offering users the option to run older agents, similar to how OpenAI provides a choice between GPT-4 and GPT-5. He also thinks Replit could have done better by setting usage caps or warnings during the rollout to prevent surprises.
What Could Replit Do Next?
While Replit’s CEO, Amjad Masad, acknowledged the problems and said the company is actively working to fix them, there’s no word yet on specific measures like credits refunds. Andersen from Moor Insights predicts that Replit probably won’t roll back prices but might offer some post-hoc credit adjustments, as other AI tool providers have done to ease user frustrations.
Replit is caught between rising costs for AI models and investor pressure to keep prices competitive. The company recently raised $250 million in funding from big names like Andreessen Horowitz and Y Combinator, around the same time it launched Agent 3. As users continue to voice their concerns, Replit and other vibe coding tools will need to find a balance between providing powerful AI and keeping costs manageable for their users.
In the broader AI landscape, companies like GitHub are also shifting to usage-based billing for their AI tools, signaling a move towards more transparent and scalable pricing models. Replit’s next steps will be crucial in maintaining user trust while managing the high costs associated with large language models.
What do you think?
It is nice to know your opinion. Leave a comment.