Hello, I would like to express my sincere gratitude for my recent experience using your tool. The reason is simple. I had never used Cloud Opus the way I’m using it now — mainly because it used to be extremely expensive, and as a beginner programmer with limited resources, it was simply out of reach for me.
However, with Cursor’s new pricing model, I’ve been able to accomplish so many tasks using Cloud Opus — and the best part, without any additional costs. I’m genuinely happy and wanted to share this joy with the community.
One thing I noticed tho. Sonnet-4.0 seems to not have the thinking activated anymore, even tho I use it with Thinking. This happened after a while. Anybody else experienced this? I never see the thoughts now. Despite using this:
true! at beginning of chats it does work still. somehow once conversation is too long, it doesnt show the thinking process anymore. hard to tell if its more noob or not maybe it does use the thinking in background and just not show it.
But if you use Claude 4 sonnet MAX. You will see that it does not have this problem, or still has it if the conversation is too long, but the length is increased many times compared to the normal Claude 4 sonnet. If you need a powerful Agent similar to Claude 4 sonnet but still powerful even when the conversation is very long, that is Claude 4 Opus (very expensive)
But in short, Claude’s 4 normal sonnets are already great. Better than 3.7 sonnets, 3.5 sonnets. We should cherish this, not “get the elephant and ask for the ivory”.
If you turn off the in-use charging and use the new cursor model, you will see that when you reach the maximum you will receive a warning after a few hours of recharging. I’ve been working all day since I made this post until now Every time a model’s Max limits me I jump to the next one lol And fantastic
@proteus-dev I’m using max too and your answer is inaccurate, OP seems to be using opus thinking max and is still within his own limits. The interpretation that the user sent 15requests is false, its 15 billable events (anything where tokens are consumed). I see very similar patterns with Claude 4 Sonnet Thinking Max.
Bro I have been using almost 100 requests with Claude 4 Opus since morning with this MCP Server tool. Really saving my rate limit even though Claude 4 Opus is very expensive. The new Pro package is really great.
I’m saying that that is a real looking consumption of ONE chat thread as I see similar stats. I had similar output with Max also before the change to new plan. it just said 0.9 or 2.5 requests, for each line. Max is counted by token usage of each part of the request, not when one request goes from begin to end, but each step consuming tokens.
Can someone enlighten me? Opus und max was always Pay-as-you-go, so maybe there is my knowledge gap? Now its included in pro for „free“?
And also: even if this was valid, it would only further highlight the huge imbalance regarding how users are provided resources. He got lucky. Good for him. Now compare with orher posts, where people can fire 5 non-max non-opus prompts (arguably far less compute) and get rate limited.
Cursor on pro is birderline unusable right now. It became a paid demo.
It has more to do with token management. While I do not know specifics of other cases I see from my own usage based pricing usages that when i dont put unnecessary content into request and have a focused task the cost is often very little.
maybe it does use the thinking in background and just not show it.