Trending Articles
Trending Topics
Trending
Anthropic has quietly restructured its Claude enterprise billing, bringing an end to the unlimited AI buffet that customers were previously enjoying.
According to a report in The Information, Anthropic has moved its largest business customers from per-seat fees to per-token pricing, ensuring that customers now pay for how much compute they consume. Organisations still on per seat plans will be migrated to per-token pricing when their contract ends, the report claims.
An update to the Claude Enterprise plan help pages stipulates in bold that “Usage isn’t included in the seat fee”, but merely gives each user access to Claude on the web, desktop and mobile apps, as well as Claude Code and Cowork. “Every token your team uses – in chat, Claude Code, or Cowork – is billed at standard API rates on top of your seat cost.”
The shift in enterprise charging structure is yet another sign of Claude’s growing pains and its struggle to provide reasonable levels of usage to paying customers. Claude saw a huge surge in demand earlier this year, after Anthropic stood up to the US government’s demands to remove safety guardrails and OpenAI moved in to take its place.
The company announced earlier this month that its annual run-rate revenue had exceeded $30 billion, up from $19 billion in early March. However, revenue doesn’t always translate to profits, particularly in the hugely expensive AI business, meaning the increased business may actually be costing the company more money. Hence the need to tighten usage limits.
Customers on Claude’s £13/$17 per month Pro plan have recently been complaining about usage limits being much more restrictive. “Typically I’ll ask it for 2-3 reports before I go to work so I can read them when I get home,” said one Pro subscriber on Reddit.
“This usually eats up about 18% of my usage limit for that day/window, then I’m able to ask Claude questions about the report without hitting any limits. Today I did my usual thing, and hit my daily limit with just the third report.”
In my own tests with Claude Pro, I’ve managed to hit session usage limits after only two or three prompts that weren’t particularly output intensive, something I’ve never seen with the comparably priced ChatGPT Plus plan.
Earlier this month, the company offered me a “one-time credit” for extra usage on my account, while explaining that, starting immediately, “third-party harnesses like OpenClaw connected to your Claude account will draw from extra usage instead of from your subscription”.
Some customers have accused Anthropic of “nerfing” the performance of Claude Opus 4.6 and Claude Code, deliberately degrading the service so that it uses less compute. Anthropic staff members have publicly denied this is the case, claiming product tweaks such as switching on “adaptive thinking” by default may explain the varying quality of results that customers are experiencing.
Perhaps to head off some of this controversy, there are reports that Anthropic is preparing to launch Claude Opus 4.7 this week. The upgrade is expected to include a new tool that improves the design of AI-coded websites.
The company has also revamped the design of Claude Code in its desktop app, introducing a new sidebar for managing multiple sessions, drag-and-drop layouts and an integrated terminal for running tests while you build.
