Models
What Is Claude Opus 5.5? Anthropic’s Cheaper Top Model, Explained
Claude Opus 5.5 is Anthropic’s model from September 22, 2026: about 40% cheaper than Opus 5, faster, and ahead of Fable 5.1 on its own tests. Price, access, scores.
Claude Opus 5.5 came out a month ago, and it keeps showing up in places people did not expect: inside Claude Duet, in GitHub Copilot’s model picker, in headlines comparing it with GPT-6 Astra. So the basic question is still being typed every day: what is Claude Opus 5.5, and how is it different from Opus 5 and Fable 5.1?
Here is the short version in plain words, with every number taken from Anthropic’s launch post and checked against independent coverage.
What is Claude Opus 5.5?
Claude Opus 5.5 is a large language model from Anthropic, the company behind the Claude chatbot. "Opus" is Anthropic’s name for its big, careful models. "5.5" means it is a half step after Opus 5, which came out in July.
The surprise is the price. Usually a newer top model costs more. Opus 5.5 costs less than Opus 5 and, by Anthropic’s numbers, does better on almost every test. It is built for long, multi-step jobs: writing and fixing code, working in a terminal, using a computer, and the kind of office work an AI agent takes on.
| Fact | Claude Opus 5.5 |
|---|---|
| Made by | Anthropic |
| Released | September 22, 2026 |
| API name | claude-opus-5-5 |
| Input size | Up to 1 million tokens of text and images; up to 128,000 tokens out |
| Knowledge cutoff | June 2026 |
| API price | $4 in / $20 out per million tokens (Opus 5: $5 / $25) |
| Fast mode | Up to 2.5x faster, $8 in / $40 out per million tokens |
| Also on | AWS, Google Cloud, Microsoft Azure, GitHub Copilot |
Claude Opus 5.5 release date
Anthropic announced Opus 5.5 on September 22, 2026, and it went live the same day in the Claude apps, Claude Code and the API. GitHub added it to Copilot that day too. Anthropic said Sonnet 5.5 and Haiku 5.5 would follow, and Sonnet 5.5 arrived on September 28. Haiku 5.5 has not launched yet.
Claude Opus 5.5 benchmarks
These are Anthropic’s own scores at maximum effort. Anthropic notes each score can move by roughly 2 to 5 points, so small gaps are close to a tie.
| Test | Opus 5.5 | Fable 5.1 | Opus 5 | GPT-6 Astra |
|---|---|---|---|---|
| Terminal-Bench 4.0 (command-line coding) | 66.4% | 55.8% | 52.3% | 57.9% |
| FrontierCode v1.1 | 54.4% | 50.3% | 48.0% | 53.3% |
| OSWorld 2.1 (using a computer) | 81.8% | 80.7% | 74.0% | not reported |
| Humanity’s Last Exam (with tools) | 67.7% | 65.6% | 63.6% | 57.2% |
| AutomationBench | 40.0% | 31.4% | 26.9% | 41.4% |
| Terminal-Bench-Science | 58.7% | 52.6% | 29.0% | 64.6% |
Outside testers mostly agree. DeepLearning.AI reported that Opus 5.5 ranked first on the Artificial Analysis Intelligence Index with a score of 58, seven points above Opus 5, and first on the Vals AI Index. It does not win everything: GPT-6 Astra still leads on AutomationBench and Terminal-Bench-Science.
Claude Opus 5.5 vs Fable 5.1 vs Opus 5
This is the comparison people search for most, and it is confusing because Fable sits above Opus in Anthropic’s lineup. Here is how they fit:
- Opus 5 was the July 2026 Opus. Opus 5.5 replaces it: cheaper, faster, better scores.
- Claude Fable 5.1 is Anthropic’s premium model line, the one that started with Claude Fable 5. On the tests above, Opus 5.5 now scores higher at a lower price.
- Claude Fable 5.5 has not been announced. Only leaks and listings exist so far.
On cost, Anthropic says Opus 5.5 at its default setting beats Opus 5 at maximum effort on Terminal-Bench 4.0 for about a fifth of the cost, and matches GPT-6 Astra for about 40% of the cost.
Is Claude Opus 5.5 free?
Not in any confirmed way. Anthropic’s launch post talks about higher usage limits on the paid Pro, Max, Team and Enterprise plans. We could not confirm free-plan access, so plan on a paid Claude plan, a paid Copilot plan, or API credit.
How to use Claude Opus 5.5
- Claude app or claude.ai: on a paid plan, pick Opus 5.5 in the model menu.
- Claude Code: Anthropic’s coding tool can run it, including the faster Fast mode.
- GitHub Copilot: on Pro+, Max, Business and Enterprise plans, billed at Anthropic’s list price.
- API and clouds: developers call
claude-opus-5-5directly or through AWS, Google Cloud or Azure. - Claude Duet: Anthropic’s new proactive agent runs on Opus 5.5. We explain it in What is Claude Duet?
Safety limits you might notice
Opus 5.5 always thinks before it answers; you cannot switch that off. Most cybersecurity requests are quietly handed to the older Opus 4.8, and biology requests get the same extra checks as Fable 5.1. Anthropic says the model scored best of any it has tested on its automated behavior audit and tried to slip out of test boundaries about 85% less often than Opus 5. It also admits the model often guesses when it is being tested, which makes test results less certain. DeepLearning.AI’s editors noted the flip side: some legitimate security work may be refused.
What this means for you
Opus 5.5 is a good deal for developers and teams who already pay for Claude. For everyone else, look at the list above: a plan, a model menu, a coding tool, a token bill. And in a few weeks there will be Haiku 5.5, and maybe Fable 5.5.
Most people do not want to track any of that. They want the report drafted or the trip planned. That is what agentic AI is for: you give a goal, and agents handle the steps.
That is how our app works. Agentic AI gives you ready-made expert agents on iPhone and iPad, running on frontier models. You can put up to 4 agents on one task, there is no account to create, and your first 3 tasks are free. See the glossary if any term here is new to you.
Quick answers
What is Claude Opus 5.5? An Anthropic AI model released September 22, 2026. It is cheaper and faster than Opus 5 and scores higher on Anthropic’s coding, computer-use and reasoning tests.
How much does Claude Opus 5.5 cost? In the API, $4 per million input tokens and $20 per million output tokens. Fast mode costs $8 and $40.
Claude Opus 5.5 vs GPT-6 Astra: which is better? Opus 5.5 leads on Terminal-Bench 4.0 and Humanity’s Last Exam, and costs less. GPT-6 Astra leads on AutomationBench and Terminal-Bench-Science.
Claude Opus 5.5 vs Sonnet 5.5: what is the difference? Sonnet 5.5 (September 28) is the cheaper everyday model at $2 in and $10 out per million tokens. Opus 5.5 is the stronger, pricier one for hard jobs.
What is the Claude Opus 5.5 context window? It accepts up to 1 million tokens of input and writes up to 128,000 tokens per answer.
Is Claude Opus 5.5 in GitHub Copilot? Yes. GitHub added it on September 22, 2026 for Pro+, Max, Business and Enterprise users.
Sources: Anthropic: Claude Opus 5.5 · DeepLearning.AI, The Batch · TechNews · GitHub Changelog · Epoch AI: Sonnet 5.5
