AI Updates · 8 min read
Claude Opus 5.5: The Interesting Part Isn't Just the New Model
Claude Opus 5.5 is Anthropic's first Claude 5.5 model. Anthropic says it performs at the level of Claude Fable 5.1 on most work while costing about 40% less to run than Opus 5.
"Key Takeaways"
- Anthropic launched Claude Opus 5.5 on September 22, 2026 as the first model in the Claude 5.5 family.
- Anthropic says it performs at the level of Claude Fable 5.1 on most work and costs about 40% less to run than Opus 5 on typical workloads.
- API pricing is $4 per million input tokens and $20 per million output tokens, with cache reads at $0.20 per million.
- Anthropic says output is more than 30% faster than Opus 5, and that Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks.
Summarize this article with
A new Claude model is easy to announce. A cheaper path to near-frontier work is harder to ignore.
Claude Opus 5.5 is Anthropic's first Claude 5.5 release. The interesting part is not only what it can do. It is what Anthropic says you pay, and how much compute the same job uses.
Claude Opus 5.5 in plain English
Opus 5.5 sits at the top of Anthropic's Claude AI lineup for this release wave. Anthropic says it performs at the level of Claude Fable 5.1 on most work, while requiring less compute to serve than Opus 5.
Anthropic also says the model was tested before release by external evaluators, including Frontier Design and METR, and that it scores strongly on Anthropic's own automated behavioural audit. Those are Anthropic-reported evaluation claims, not a substitute for your own review process.
What changed from Opus 5
Anthropic frames Opus 5.5 as a major step up from Opus 5 on complex work: long coding jobs, codebase migrations, agents, and professional knowledge work. Early-tester stories in the announcement include large migrations and audits finished in far less wall-clock time than Opus 5.
Anthropic also says communication is clearer: more natural writing, important points first, easier to follow over long sessions. If Opus 5 felt stiff or hard to skim, that is the everyday change to test.
The 40% cost story, from Anthropic
| Meter | Claude Opus 5.5 | Claude Opus 5 |
|---|---|---|
| Input tokens / 1M | $4 | $5 |
| Output tokens / 1M | $20 | $25 |
| Cache reads / 1M | $0.20 | $0.50 |
| Cache writes / 1M | $5 | $6.25 |
Why token efficiency matters for agents
Agent loops reread the same files and tool results again and again. Anthropic says cache reads make up a large share of agentic and coding-work costs, which helps explain why the lower cache-read price matters.
Anthropic says cache reads are 60% cheaper than on Opus 5, and that Opus 5.5 also uses fewer tokens per task. Together, Anthropic reports about a 40% drop in cost on typical workloads, plus output that is more than 30% faster than Opus 5.
Coding, communication, and safety claims
On coding, Anthropic highlights long and sprawling jobs: migrations, audits, agentic terminal work. It publishes strong scores on suites such as Terminal-Bench 4.0, FrontierCode, and CursorBench. Treat those charts as Anthropic-reported, with Anthropic's own footnotes about effort settings and safeguard interventions.
On safety, Anthropic says Opus 5.5 is the strongest model it has tested on its automated behavioural audit, and that it is deploying safeguards similar to those used for Claude Fable 5.1 in biology and cybersecurity domains. Vetted organisations can apply to life-sciences and cybersecurity verification programmes, according to Anthropic.
Anthropic also notes that at these capability levels, benchmark margins are a less reliable guide to real-world differences, and that the gap with Fable 5.1 can feel narrower than the scoreboard suggests. That honesty is useful. Your repo is still the exam.
Who should care about Anthropic Opus 5.5
- Teams already paying for Opus or Fable-class work who want similar quality at lower agent cost.
- People running long coding or research agents where cache traffic is the silent bill.
- Writers and operators who found Opus 5 capable but awkward to read over long sessions.
- Anyone waiting for the rest of Claude 5.5: Anthropic says Sonnet 5.5 and Haiku 5.5 follow in the coming weeks.
Frequently Asked Questions
What is Claude Opus 5.5?
It is Anthropic's September 22, 2026 flagship release and the first model in the Claude 5.5 family. Anthropic says it reaches Fable 5.1-level performance on most work while costing less to run than Opus 5.
How does Claude Opus 5.5 vs Opus 5 pricing compare?
Anthropic lists Opus 5.5 at $4 / $20 per million input and output tokens, against $5 / $25 for Opus 5, with cache reads at $0.20 versus $0.50. Anthropic says typical workloads cost about 40% less.
Should I switch from Fable 5.1 immediately?
Only after you re-test your own long jobs. Anthropic says the everyday gap can be narrower than the charts. Price and speed may matter more than another crown.
The Chai Takeaway
If your Claude bill is mostly agents rereading the same codebase, Opus 5.5 is worth a careful cost test this week. If you only need short answers, wait for Sonnet and Haiku 5.5. The story here is useful capability getting cheaper to leave on.
Explore More AI Updates