I've been really impressed with Claude Opus 4 after it managed to solve a complex 9000-word quantitative finance problem that had stumped me for weeks. Unlike earlier versions like o3 and o1-pro, it was able to understand the deeper nuances of the problem and suggested a relatively straightforward solution I hadn't thought of despite hours of deliberation. Now, I'm considering switching to the Claude Max plan, but I'd love to hear about people's experiences with Opus 4 and any rate limiting they encountered. Do you hit limits on the $200 plan frequently, or is that just a precaution against misuse?
7 Answers
If you want to test the waters without diving into the Max plan, I have an extra slot on my Team plan available for $30 a month, and I rarely hit any limits. I used it heavily during a deep work session recently and had no issues at all!
I'm not a pro, but I can give you a rough estimate. From what I've gathered, on the Max plan, you might get around 45 to 180 messages for Opus within that 5-hour period. It really depends on how long your messages are and the overall conversation length. Claude Pro users have a significantly lower message limit, and the Max plan boosts that quite a bit—but keep in mind Anthropic's limits aren't super clear.
I've tried using it with my Pro account, but I keep getting 'invalid model' errors, so I'm not sure what's going on there.
From what I've seen, I hit the Opus 4 limits in about 2-3 hours while using the Claude Max $200 plan, but keep in mind that I run multiple agents—like 8 or more—at once. If you're just a regular user, you might get about 5-6+ hours before hitting a rate limit.
I heard that one session is considered a 5-hour block, and if you exceed 50 sessions a month on the Max plan, they'll cut it off. Just going off memory but it's something like that.
Haha, what kind of projects are you working on that you need so many agents?
Honestly, the small context window size of Claude Opus 4 is a bit of a letdown. I expected it to be much larger than just 200K tokens, and it gets expensive really fast. Just a couple of prompts and I'm already waiting hours for more. Compared to o3, it really falls behind. It's great if your codebase is small, and it follows rules well, but for anything more complex, it’s frustrating.
I don't get why Anthropic is focusing heavily on coding when there are still major quality of life issues with Claude. I'm used to Grok and ChatGPT, and honestly, the limits with Claude are pretty rough. That's just not enough for regular usage and even less acceptable when you see what other models offer.
I honestly don't hit the limits very often with Claude Code—maybe 3 times this past month, and I use it daily.
So far, using Claude Max has been solid for me. I manage to get a decent amount of usage out of it in testing, but I can't give exact numbers. Just a 'good amount' overall.
That doesn't quite sound right. I've been using Opus on the Pro plan, and I've sent dozens of messages without hitting limits.