Claude Opus 5.5 is Anthropic's newest flagship model, released on 22 September 2026. The headline changes are simple: extended thinking is on by default, it is cheaper and over 30% faster to run than Opus 5, and it goes much further on long coding and agent tasks without losing the thread. It launched on the Pro, Max and Team plans, with Sonnet 5.5 and Haiku 5.5 due to follow. Prices and limits move quickly, so confirm the current numbers with Anthropic before you commit.
What actually changed in Opus 5.5?
This is an efficiency-and-endurance release rather than a flashy new-feature one. Three things matter for everyday work.
Thinking is on by default
Opus 5.5 reasons before it answers, without you turning anything on. In practice that means fewer confidently-wrong first drafts and better results on problems that need a plan - migrations, multi-file edits, anything with several moving parts. You can still ask for a quick answer when you want one.
It is cheaper and faster
Anthropic set Opus 5.5 below Opus 5 on price and said it runs typical workloads more cheaply, with output generated over 30% faster and much lower cache-read costs. The direction of travel is the same one we wrote about in the AI price war: top-tier models keep getting cheaper, so the cost of putting one to work keeps falling.
It goes further on long tasks
The clearest gains are on sprawling, multi-step work. Anthropic's own examples describe auditing and fixing very large codebases in a fraction of the time Opus 5 needed, and holding quality across long agentic runs. If your work is one big task rather than a quick question, this is the upgrade you feel.
A faster, cheaper model is only worth it if you point it at real work. William is the role already written - inbox triage, scheduling, follow-ups and briefing docs - so Opus 5.5 has a job the moment you switch it on. Runs on Claude.
View William →How to get the most out of Opus 5.5
A better model rewards a better setup. Four habits do most of the work.
- Give it a role, not just a prompt. A defined role with a working method beats re-explaining the job every session. That is the whole idea behind a skill versus a prompt - reusable context the model loads automatically.
- Keep autonomous runs checkpointed. Opus 5.5 can run for a long time on its own, which is powerful and risky. Break big jobs into stages with a human checkpoint rather than one heroic overnight run, so an early mistake does not become load-bearing.
- Put it where the work already lives. The model is most useful next to your files, your repo or your inbox. Integration beats raw capability more often than people expect.
- Reuse context instead of re-prompting. A clear system prompt or a saved skill saves tokens and gives steadier output than pasting the same background in again and again.
Should you switch?
If you are already on Opus 5, moving to Opus 5.5 is close to a free upgrade - same kind of work, lower cost, faster, and better on long tasks. If you are coming from an older or lighter model, the gains are largest on complex jobs and smallest on short, simple questions. For most people the honest answer is: use it for the hard work, and a lighter model for the throwaway stuff. If you want the full picture of the previous generation, see our Claude Opus 5 guide.
Frequently asked questions
When was Claude Opus 5.5 released?
Anthropic released Claude Opus 5.5 on 22 September 2026. At launch it was available on the Pro, Max and Team plans, with Sonnet 5.5 and Haiku 5.5 announced to follow in the following weeks.
Is Claude Opus 5.5 cheaper than Opus 5?
Yes. Anthropic priced Opus 5.5 below Opus 5 and said it runs typical workloads at a lower cost, alongside faster output and lower cache-read pricing. Exact rates change over time, so check Anthropic's current pricing page before you budget around it.
What is Opus 5.5 best at?
Long, multi-step work: auditing and fixing large codebases, running agentic tasks that take many steps, and keeping quality over a long session. Extended thinking is on by default, so it reasons before answering without you having to switch anything on.
Do I need to change how I prompt for Opus 5.5?
Not really. The bigger lever is giving it a defined role and reusable context - a skill or agent - instead of re-explaining the job each time. That is where a stronger model pays off most.
Opus 5.5 is strongest on long, multi-step work - exactly what a chief of staff does. Ryan turns a messy week into priorities, decisions and next actions, with the intake questions and method built in. Install once and it loads whenever the task fits.
View Ryan →In summary: Opus 5.5 is cheaper, faster and better at long, multi-step work, with thinking on by default. The way to capitalise on it is to give it a role and reusable context rather than re-prompting from scratch. Browse ready-made AI agents and skills that run on Claude, try the free AI generators first, or get everything with KissMySkills Unlimited from $9/month.
