Anthropic tests move Claude Code out of Pro subscription, OpenAI states Codex will retain a free tier

According to a Beating monitoring report, Anthropic has taken Claude Code down from the Pro subscription, with no announcement—only changing the pricing page. If new Pro users want to use Claude Code, they must upgrade to the Max subscription starting at $100/month. After that, Anthropic’s growth lead Amol Avasare posted on X to clarify that this was a small-scale test targeting about 2% of newly registered Pro users, and current Pro and Max subscribers are unaffected.

Amol explained that when Max was launched a year ago, it did not include Claude Code, and Cowork and long-running agents did not exist either; at the time, it was designed only for heavy chatting. After Claude Code was bundled into Max and usage surged following the release of Opus 4, long-running asynchronous agents became a standard daily workflow, and average usage per subscriber rose significantly; existing subscriptions were not designed for this kind of use. Anthropic previously added a weekly cap and restrictions during peak hours, and it is still testing different approaches and collecting feedback. He said that if the final plan ends up affecting existing subscribers, they will be notified well in advance: “The message will come from us, not screenshots on X or Reddit.”

OpenAI Codex head Thibault Sottiaux then weighed in on X: Codex will continue to be provided in the Free and Plus ($20/month) subscriptions. OpenAI has the compute power and efficient models to support it, and any major changes will be communicated to the community in advance. “Transparency and trust are two principles we won’t break, even if it means earning a little less in the short term.” Sam Altman retweeted Tibo’s post with the caption, “We hope you have a lot of AI.”