This feels like another moment, like earlier this year, and the year prior. I am thrilled for my use cases in the applications that I build with CC, and for the LLM API usage that I sell. However, it honestly scares the crap out out of me.
Anthropic's naming convention seems to follow an S-curve. So Opus 5 is actually a little worse than 4.7, but by the time it hits 5.5, it's at the steep part of the new S-curve.
Similar with Claude 4 being worse than 3.7, but peaking at 4.5. At 4.6 they started to hit the restrictions and 4.7 seemed to patch what they made worse.
Yes, Opus 5.5 and Astra-6 are big improvements over Opus 5 and Sol-5.6. I'm at a point where I can automate much of my workflow, even for complicated tasks, which I could never do before.
Also the new Xiaomi model, Mimo v2.6, is outstanding as a light and cheap model. For coding, the new meta is using Astra-6-Max for orchestration and architecture, and Mimo for grunt work.
Yep, I'm personally finding Opus 5.5 to be the first real leap I've felt since Opus 4.5. The time-to-first-token seems dramatically better in Claude Code compared to Fable 5.1/Opus 5 as well, which really helps both interactivity and also overall time-to-completion.
It also burns Claude subscription quota much more slowly than Fable, which is nice.
I’m also finding Opus 5.5 to be a significant advance. In the past few days, I’ve given it several complex tasks related to two long-term dictionary-building projects [1, 2], and it handled them impressively—similar to Fable but with much lower token consumption. It also communicates better than Opus 5.1; I don’t think I’ve seen any obvious Claudisms yet.
I have Opus 5.5 running subagents for the last 20h without any input from me after developing a plan and it seems to be going well somehow... but I will find out soon...
I've been checking it and it's doing great I'm sure there are some bugs but it's doing a huge bulk of tedious simulation logic for my game. I was previous in the loop quite a bit but I found myself never having to redirect it so I tried the here a giant plan /loop until its done
First of all, I consider it very expensive, somehow in the range of Fable:
I activated Opus 5.5 when the invitation popup showed up, and Im already paying as a pro-user - nontheless the usage of Opus 5.5 charged massivly additional on the last few days, around 12-13 USD per day.
Im still fine with Opus 4.6 or 4.8 for daily development.
It's the first model that I feel I can just use freely without crashing into the 5 hour limits. I just manage to get my work done and still have tokens left over
Similar with Claude 4 being worse than 3.7, but peaking at 4.5. At 4.6 they started to hit the restrictions and 4.7 seemed to patch what they made worse.
Also the new Xiaomi model, Mimo v2.6, is outstanding as a light and cheap model. For coding, the new meta is using Astra-6-Max for orchestration and architecture, and Mimo for grunt work.
It also burns Claude subscription quota much more slowly than Fable, which is nice.
[1] https://www.tkgje.jp/
[2] https://tkgally.github.io/eex-dict/index.html
at that point the hard part is just trusting it didnt quietly go off track somewhere imo
I activated Opus 5.5 when the invitation popup showed up, and Im already paying as a pro-user - nontheless the usage of Opus 5.5 charged massivly additional on the last few days, around 12-13 USD per day.
Im still fine with Opus 4.6 or 4.8 for daily development.