Claude Opus 5.5 costs 40% less than Opus 5: what improves and what to check right away in Claude Code

Claude Opus 5.5 costs 40% less than Opus 5 on a typical workload. Anthropic released it on September 22, 2026, and the savings come from two places: the API is cheaper, and the model burns fewer tokens to reach the same result.

Why does Claude Opus 5.5 cost less than Opus 5?

The first reason is the price list. According to Anthropic’s announcement, Opus 5.5 costs:

• $4 per million input tokens, versus $5 for Opus 5;
• $20 per million output tokens, versus $25 for Opus 5.

The second reason is efficiency. Opus 5.5 gets to the same result with fewer steps and fewer tokens. In the announcement, Factory reports that Opus 5.5 at medium effort matched Opus 5 at high effort while using 20 to 25% fewer output tokens. Deloitte reports that Opus 5.5 at its lowest effort setting caught 72% of known bugs in code reviews, versus 56% for Opus 5 at high effort.

In practice: with Opus 5.5 you get the same quality of output and pay for fewer tokens. A lower price per token and fewer tokens per task: that is where the 40% comes from.

How does Opus 5.5 compare to Fable 5.1 and Mythos 5.1?

Opus 5.5 performs close to Anthropic’s most powerful models. According to Anthropic, Opus 5.5 works at the level of Claude Fable 5.1 on most tasks and is comparable to Claude Mythos 5.1 in biology and cybersecurity.

The difference is the price. On Anthropic’s official pricing page, Fable 5.1 and Mythos 5.1 cost $10 per million input tokens and $50 per million output tokens: more than double Opus 5.5. Mythos 5.1 also has limited availability.

What does Claude Opus 5.5 do better?

Opus 5.5 adds no headline features. It does the same things as the previous models, only better. Three improvements matter if you use it every day.

It handles long tasks without losing track

Opus 5.5 stays on task even during long sessions on very large codebases. In the announcement, one tester describes completing a 680,000-line code migration in less than a day. Clio reports that the model stayed focused for over 18 hours on a task spread across multiple repositories.

It got better at visual work

In one test, several Claude models built a game from a single prompt. Opus 5.5 produced the best result for graphics and polish.

It communicates better while it works

Opus 5.5 was trained to narrate its work as it goes: what it is doing, what it found, and what it needs to move forward. Anthropic says its messages put the most important information up front, and testers say this helps during long working sessions.

What should you check in Claude Code after switching to Opus 5.5?

If you use Claude Code, two changes that shipped in version 2.1.280 are worth checking right away.

1. Your saved effort is not inherited. The effort level you set on previous models does not apply to Opus 5.5, which starts from its own default. Run /effort and check which level you are on.
2. On Pro plans the default model is now Opus. Claude Code moved the default model on Pro and Team Standard plans from Sonnet to Opus. If you used Sonnet before, you may hit your limits sooner, because Opus uses more of your quota. Run /model and pick the model you actually want.

Both changes are listed in the Claude Code changelog. Opus and Sonnet also behave differently in Claude Code in other ways: I covered that in Opus 5 too verbose in Claude Code.

So what does this mean if you use Claude every day?

Opus 5.5 brings top-tier model quality at a lower price, with fewer tokens wasted on long tasks. The savings only show up if your effort level and model are set correctly. Before your next long task in Claude Code, run /effort and /model: two commands that decide how much you pay for every result.