Anthropic pulled a fast one on us! (Opus plans LIMITED)

Summary

Anthropic recently made a quiet change to its subscription plans that directly affects how much you can get out of your paid tier. If you're on the Free, Pro, or Max plan, you're now burning through your session tokens significantly faster than before — especially during peak hours. Anthropic did not announce this change upfront, which frustrated many users who initially thought it was a glitch. It's not. It's intentional. The most likely reason is that Anthropic needs to redirect compute resources toward training its next model, Mythos, which means less subsidized usage for everyday subscribers. API users are unaffected since they pay per token anyway, but if you're on a flat monthly plan like the $20 Pro or $200 Max tier, you're getting less value for the same price. The Max plan users are hit hardest since they tend to be the heaviest users. One additional frustration is that Anthropic, unlike OpenAI, Google, or even Chinese model providers like Minimax, does not publicly publish its token limits or rate limits — giving them flexibility to make these changes silently. The video also touches on alternatives like GLM5 and Minimax 2.7, which are cheaper and increasingly competitive, especially for agentic use cases. However, even those alternatives are experiencing strain due to surging AI demand. The key message: if you rely on Claude for heavy coding or agentic workflows, your experience has quietly gotten worse, and now may be a good time to explore alternatives — though no perfect substitute exists yet.