Is Claude Max $100 actually worth it compared with Codex?
retained text
This week I tried the $100 Codex plan for the first time after all the marketing around Astra. I normally use the Plus/Pro tiers on both platforms, and until now I've been able to build my projects pretty comfortably with both. When Astra came out, I actually liked it quite a lot. I'm also starting to work much more seriously with AI now, as I'm building larger and more complex projects. My background is mainly product design, and graphic design is probably my strongest area, so I'm quite demanding when it comes to frontend work. Astra was good for the first few days, but this week it has been so unbelievably bad that I ended up requesting a refund. I've wasted a ridiculous number of hours because it simply doesn't follow instructions. You ask it to change one thing and it changes absolutely everything. When it comes to frontend, I think it's genuinely bad. Not necessarily because the output screams "AI design" as much as it used to. I actually think it has improved somewhat in that regard. The problem is that it seems to have one predefined design template that you just cannot get it away from. It constantly produces the same thin-arrow buttons, buttons with bottom borders, huge typography, square images, the same kinds of layouts, etc. The first project it generates, you might think: "Wow, this has improved." By the third project you're thinking: "Please, not this again." And no matter how much I try to guide it toward a properly considered design direction, I can't seem to get it there. Since I expect the refund to go through, I've been considering Claude Max instead. The problem is that Opus 5 frustrates me even more. For my use case, Opus 5 is honestly one of the worst models I've tried. It doesn't seem to matter what task I give it, I very rarely like the result. I currently have Claude Pro and I don't even reach 30% of my weekly usage because I simply don't use Opus much. I don't like the way it responds either: it talks far more than I ask it to, and for frontend design it's probably the worst model I've personally used by quite a distance. I also find that it doesn't follow instructions particularly well. However, when I look at comparisons between Fable 5.1 and Astra, Fable looks much better to me. Better design, more original, and generally more interesting. My main concern is usage limits. I've become very used to not having to think about a five-hour usage window. If Fable can handle the heavy lifting and still leave enough usage for me to implement and iterate with Opus, that could be ideal. This month I also tried Grok 4.6 inside Cursor, and I liked it significantly more than Opus. It's very direct, follows instructions well, and doesn't constantly try to do more than what you asked. So Claude + Grok doesn't seem like a bad combination to me. What worries me is the limit. With Opus 5, two messages on a large implementation can sometimes burn through a huge amount of my allowance. I really don't want constant interruptions in the middle of working. For those of you on Claude Max: Have you also noticed a drop in quality recently, similar to what people have been saying about Astra? And realistically, how restrictive is the $100 Max plan if you're using it heavily for larger coding projects? The other thing making the decision harder is that I find the Codex app significantly better than Claude Code as an actual working environment. Going back would be a bit painful.
document fetched 2026-09-21 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1wlk92z/is_claude_max_100_actually_worth_it_compared_with