Over the past two weeks, ChatGPT Pro usage limits seem to have decreased significantly.
On August 5, I used my personal usage reset and then consumed 195.5 million tokens (5.6 Sol High - xHigh), based on the usage data shown in my ChatGPT account on August 6. According to the usage page, this represented 28% of my weekly allowance. Based on that figure, my total weekly limit would be approximately:
195.5M ÷ 0.28 ≈ 700 million tokens
For comparison, my colleague has a $100 Claude subscription and uses it for almost the same type of work. Based on our token-usage calculations, using only Opus 5 in xhigh mode, Claude currently appears to provide more than twice the usage capacity—approximately 1.5 billion tokens per seven-day period.
I understand that OpenAI and Anthropic may account for tokens differently, particularly when cache reads, cache writes, and output tokens are involved. Therefore, this may not be a perfect one-to-one comparison. However, the difference is still substantial.
I have also seen many recent posts from Pro users reporting that their available usage has dropped dramatically over the past two weeks. This suggests that the issue is not limited to my account and may reflect a broader reduction in Pro usage limits.
The ChatGPT interface and overall user experience are significantly better for me, and I would prefer to continue using it. However, if these limits remain unchanged, I will almost certainly switch to Claude when my current ChatGPT subscription ends later this month. At the same price point, the difference in available usage is becoming too large to justify staying.
Appendix: Previously, while using GPT-5.5 in xhigh mode, the limits were generous enough that some of the complimentary usage resets provided by OpenAI expired without me ever needing to use them. The contrast with the current limits is therefore especially noticeable.