What Happened
This week, screenshots from prominent tech blogger George Pu revealed that Anthropic had quietly updated its pricing page, removing Claude Code — a terminal -based AI tool that helps developers write and debug code — from the $ 20/ month Pro plan and making it exclusive to the Max plan, which starts at $100/month. That 's a 5x price increase. No email notice . No official announcement. Just a revised pricing comparison table.
Under public pressure, Anthropic executive Amol A vasare stepped in to clarify: this was a pricing test targeting 2% of new users, and existing Pro and Max subscribers would not be affected. But he also acknowledged that since Opus 4 launched, user behavior has shifted from casual chat to long-running code tasks and automated workflows — Agents , meaning AI programs that can autonomously execute multi -step tasks. The compute costs, he admitted , can no longer be absorbed by the existing plan structure , and adjust ments are coming .
Industry Reaction
Open AI moved fast . Sam Altman himself posted a pointed qu ip, and Open AI staff quickly announced that Codex would remain available on the $20 Plus plan — fr aming the whole episode as a stand for "fair ness and transparency." The PR move was well -executed , but deser ves some skept icism: Open AI's own low -price strategy is equally subsid ized by heavy capital burn , and its long -term sustainability is just as uncertain.
The more pointed criticism came from tech comment ator Ed Zitron, who asked directly : if this really was a small 2% test, why were the official help documentation and pricing comparison tables updated simultaneously ? His read is that Anthropic intended to roll out the price h ike across the board and only reversed course after being caught early . That 's not an unre asonable interpretation — over the past two months, Anthropic has quietly reduced API call quot as for several third-party tools, and Opus 4.5's token consumption ( the unit by which AI measures and charges for text processing) has increased not iceably compared to the previous generation. A number of Pro users burned through their entire weekly quota by Wednesday .
Our take : the " limited test" explanation is plausible, but so is the prediction that a broader price increase is inevitable . The two are not mut ually exclusive.
What This Means for Regular People
For enterprise IT: Teams that have built internal development workflows around Claude Code need to reassess vendor lock-in risk . AI tool pricing can change without notice at any time — workflow design should preserve the ability to switch.
For individual developers : Heavy users of AI coding tools can afford to watch alternatives like Open AI Codex and Cursor in the short term, but should recognize that competitors' low prices are not long -term commitments either .
For the broader market: Chinese large model products — such as De epSeek and Tong yi — offer lower prices and more generous usage quot as, and some overseas developers have already begun migrating to test them . This is a window of opportunity for Chinese vendors , but whether they can meet the demands of professional developers remains an open question of product maturity.