Anthropic appears to be quietly redrawing the line around Claude Code in the Claude Pro plan. New Pro subscribers are now seeing a version of the plan that does not include the coding tool, while current Pro and Max customers keep their access intact. That split sounds like a small experiment, but the pricing page suggests something more permanent is being tested in public.
The company says the change is an A/B test affecting about 2% of new signups, with the stated goal of easing server load. That is a familiar move in AI right now: usage spikes are expensive, and vendors are increasingly willing to trim features, throttle access, or invent new tiers rather than eat the cost. Google already did a version of this by carving out a separate AI Plus plan, and Anthropic may be watching that playbook closely.
What changed on the Claude Pro pricing page
The odd part is that Anthropic’s public messaging and its pricing page are not telling quite the same story. On one hand, the company says most new users should still get Claude Code. On the other, the Pro plan page now shows the feature unchecked, which is a pretty blunt signal to anyone about to hand over money. If you are a new buyer, ”maybe you’ll get it” is not exactly a confidence-building sales pitch.
That mismatch matters because Claude Code is one of the reasons some people subscribe to Pro in the first place. Take away the coding tool and the plan starts looking less like an all-access AI subscription and more like a quieter chat product with a nicer label.
Anthropic’s server-cost problem is getting visible
Anthropic is not the first AI company to discover that popular features become financial liabilities when everyone uses them at once. Coding assistants are especially hungry: they produce long outputs, hold state across back-and-forth prompts, and tend to attract power users who hammer them all day. In other words, they are exactly the kind of feature that looks generous in marketing and annoying on a balance sheet.
That is why this test is more revealing than a simple product tweak. If Anthropic keeps Claude Code behind Pro for only some users, it could be testing whether enough people still subscribe without it. If the answer is yes, expect a cleaner tiering strategy. If the answer is no, the company may quietly reverse course before too many new customers notice they bought a plan with a footnote-sized asterisk.
A new Claude tier could be the real plan
The most plausible outcome is not a simple yes-or-no switch for Pro. It is a new middle tier, positioned between Free and Pro, that strips out Claude Code and maybe a few other premium features. That would let Anthropic segment heavy users from casual ones without forcing everyone into the same expensive bucket.
For now, though, the company has created the exact kind of confusion subscription users hate: existing customers are safe, new customers may not be, and the sales page is making a stronger claim than the company’s explanation. Expect that to get cleaned up fast if Anthropic wants to avoid training people to read pricing pages like legal contracts.
The bigger question is whether this becomes the new normal for AI subscriptions. Once a headline feature starts costing too much to hand out freely, the industry rarely gives it back. It just finds a friendlier way to charge for it.