# Who actually believes Opus 5.5 is cheaper? Here is a 3x consumption anomaly, think of it what you will. — retained record

> Retained reddit record from reddit:ClaudeCode.

Canonical: https://highsignal.app/data/records/3f4bbf034de56963

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeCode)  published 2026-09-23 14:15 ingested 2026-09-24 02:40

[original source ↗](https://www.reddit.com/r/ClaudeCode/comments/1wo77mf/who_actually_believes_opus_55_is_cheaper_here_is/)

## retained text



```
I am a power user of 1 year, using 2 max20 accounts over thousands of hours. I know what to expect. My workflows are highly refined. My workspace is incredibly mature.

I could have set my thinking down a level or two, but the point here is a comparison against Fable 5.1 Parent (xhigh) + Opus 5.0 subagents (max) vs Fable 5.1 Parent (xhigh) + Opus 5.5 subagents (max).

I will share two photos. One is my usage after performing the usage reset yesterday around 2pm EST, so 20 hours ago. The other is my usage via ccusage.

Now I will stop typing and allow you to use your eyes and brain in order to reason over what I am looking at this morning.

https://preview.redd.it/m91k5zzsz9rh1.png?width=1028&format=png&auto=webp&s=ea539ae63efb49c549385df5de6839245757228b

https://preview.redd.it/gujad0323arh1.png?width=1439&format=png&auto=webp&s=96fab9acb2ba96fb1bf8a6b6021a10fb77f2052b

I could take many many photos like this, where the usage over months and months is in the range that you see it. The anomaly is September 22nd. When trying to consider what you're looking at here, think in terms of anomalies and think about what changed.

Some differences I noticed in general - Opus 5.5 agents seem to deploy up to 20% more agents. Though this does not attribute much here as we are not looking at a 20% increase in my usage. We are looking at 3x+ my typical usage (that $465 is residual from two different task that ran overnight, though during the day I used my second account for 1 task which likely cost roughly $150. Full transparency). What is truly crushing usage is the Max thinking. **Though my point is - if "Opus 5.5 is so much cheaper" then we would expect to also see that cheaper price around max thinking. At the very least we would not expect to see a massive increase. This is my point with this thread.**

The tasks performed are normal routine development as I have been doing for many, many months. **Nothing was particularly different about 9/22/2026. The only difference was Opus 5.5**

Take from this whatever you will. Yes, obviously I will be dropping to xhigh, and surely there will be some droolers that comment this. Though this is not the point. Personally, I question the actual price drop. **If we look at the benchmark graphs on this page -** [**https://www.anthropic.com/claude-opus-5-5**](https://www.anthropic.com/claude-opus-5-5) **- the natural conclusion is that in most instances you would expect to pay less as evidenced by their benchmarks. However, that is not quite the reality of the situation.** Perhaps you would not expect the full 40% decrease, **but you also would not expect a 300% increase.**

Is the model great? Yes! Certainly is. Though is it cheaper? No, not in my experience.

  
(A secondary aspect to this is a behavioral change in Fable 5.1 which seems to promote more consumption of context - whether good or bad I do not know as I have yet to get an implementation complete since 5.5 release. One element of this seems to be higher reliance on 5.5 agents where the 5.5 agents run for longer, and return a large in-line response. My process treats subagent output as leads, and with increased ability of 5.5, this may also mean more leads that need verification. 50% consumption on Fable in a single day is unusual for me. There was a need for compaction and handoffs more often, which indicates to me that my more capability = more leads theory could be close to reality. I have yet to actually dissect the sessions.)
```



document fetched 2026-09-24 00:43 · canonical https://reddit.com/r/ClaudeCode/comments/1wo77mf/who_actually_believes_opus_55_is_cheaper_here_is
