Is Claude Opus 5.5 cheaper than Opus 5?

Yes. Per Anthropic's release notes, output tokens dropped from $25 to $20 per million, input tokens from $5 to $4 per million, and cached reads fell 60% to $0.20 per million, which Anthropic says works out to roughly 40% lower cost on typical work.

How much faster is Opus 5.5 than Opus 5?

Anthropic's release notes put output generation at more than 30% faster than Opus 5, and the model also scores higher on benchmarks such as Terminal-Bench 4.0, where it reached 66.4% versus 52.3% for Opus 5.

Should I use Opus 5.5 or Sonnet 5.5?

It depends on the task. Our guide to choosing a Claude model covers when Opus's extra reasoning is worth the higher price over Sonnet, and our Sonnet 5.5 coverage looks at Anthropic's faster, lower-cost model released six days after Opus 5.5.

News · Published October 5, 2026

Claude Opus 5.5: What Changed and What It Costs Now

Anthropic released Claude Opus 5.5 on September 22, 2026, calling it the strongest-performing model it has tested to date, according to Anthropic's own release notes and TechCrunch's report published the same day. If you already reach for Opus on your harder tasks, coding problems, long reports, anything you'd normally hand to a senior person rather than a junior one, this is the update that makes doing that noticeably cheaper without asking you to change how you prompt it.

What Anthropic released

Opus 5.5 is the first model in the new Claude 5.5 family, arriving before Sonnet 5.5 and the still-unreleased Haiku 5.5. Per Anthropic's own release notes, the model performs at roughly Claude Fable 5.1 level on most work while costing about 40% less to run than Opus 5. TechCrunch's September 22, 2026 report, byline Russell Brandom, quotes Anthropic describing Opus 5.5 as "the strongest-performing model we've tested to date," and notes it beats the larger Fable model on several benchmarks while also succeeding at some informal tasks where Fable struggled.

The price cut, in numbers

According to Anthropic's release notes, Opus 5.5 pricing drops across the board compared with Opus 5: input tokens fall from $5 to $4 per million, output tokens from $25 to $20 per million, cached reads from $0.50 to $0.20 per million (a 60% cut), and cached writes from $6.25 to $5 per million. Anthropic also states the model needs less compute to serve, which is part of how it's generating output more than 30% faster than Opus 5 at the same time as getting cheaper.

What the benchmarks show

Per Anthropic's release notes, Opus 5.5 scored 66.4% on Terminal-Bench 4.0, up from Opus 5's 52.3%, and 57.8% on CursorBench 4.0 against 46.6% for Opus 5. On OSWorld 2.1, a computer-use benchmark, it reached 81.8% partial completion versus 74.0%, and on the GDPval-AA v2.1 knowledge-work benchmark it scored 1846 Elo against Opus 5's 1708. One early tester, quoted on Anthropic's release page, said of the model's writing: "It writes the way I do." If you're deciding whether that jump is worth paying for on your own work, our guide to choosing a Claude model walks through when Opus's extra reasoning actually pays off over Sonnet.

New to Claude entirely?

Claude AI for Beginners Bible covers the basics this kind of model update builds on, rated 4.5 out of 5 from 303 ratings as of August 30, 2026.

Get the book on Amazon

As an Amazon Associate, this site earns from qualifying purchases.

Where you can use it, and the safety strings attached

Anthropic's release notes list Opus 5.5 as available now in the Claude apps and the Claude Platform API under the model identifier claude-opus-5-5, plus Amazon Web Services, Google Cloud, and Microsoft Azure. Pro, Max, Team, and Enterprise plans get increased five-hour usage limits for the model, per the same release notes. Because Opus 5.5 matches Fable-level capability in biology and cybersecurity tasks, Anthropic ships it under the same safeguards it uses for Fable around exploit discovery and biological-weapons uplift, framing the approach as "pacing the rate of capabilities advancement so that risk prevention has time to keep up."

What this doesn't tell you yet

This is based on Anthropic's own release notes and TechCrunch's September 22, 2026 report, not independent, hands-on testing across everyday tasks. Benchmark scores like Terminal-Bench and OSWorld measure specific coding and computer-use scenarios, not general writing or research, so your own results on a given task may land differently. At the time of that report, Anthropic had not given exact release dates for Sonnet 5.5 or Haiku 5.5, only "the coming weeks"; Sonnet 5.5 has since shipped, which our separate coverage walks through.

Questions people are asking about this update

What did Anthropic announce on September 22, 2026?: Anthropic released Claude Opus 5.5, the first model in its new Claude 5.5 family, calling it the strongest-performing model it has tested to date, according to Anthropic's own release notes and TechCrunch's report published the same day.
Is Claude Opus 5.5 cheaper than Opus 5?: Yes. Per Anthropic's release notes, output tokens dropped from $25 to $20 per million, input tokens from $5 to $4 per million, and cached reads fell 60% to $0.20 per million, which Anthropic says works out to roughly 40% lower cost on typical work.
How much faster is Opus 5.5 than Opus 5?: Anthropic's release notes put output generation at more than 30% faster than Opus 5, and the model also scores higher on benchmarks such as Terminal-Bench 4.0, where it reached 66.4% versus 52.3% for Opus 5.
Where can I use Claude Opus 5.5?: Anthropic lists it in the Claude apps and the Claude Platform API under the model identifier claude-opus-5-5, plus Amazon Web Services, Google Cloud, and Microsoft Azure.
Should I use Opus 5.5 or Sonnet 5.5?: It depends on the task. Our guide to choosing a Claude model covers when Opus's extra reasoning is worth the higher price over Sonnet, and our Sonnet 5.5 coverage looks at Anthropic's faster, lower-cost model released six days after Opus 5.5.

Keep reading

Disclosure: this page is an independent reader resource. Some links are Amazon Associates links, which means we may earn a commission from qualifying purchases at no extra cost to you.