Claude Opus 5.5: Lower Prices, Higher Limits, Clearer Writing

Anthropic released Claude Opus 5.5 on September 22, 2026, the first model in its new Claude 5.5 family. It costs $4 per million input tokens and $20 per million output tokens, 20% less than Opus 5, and Anthropic says it does most work at the level of its top-tier model, Claude Fable 5.1. It is now the default model in Claude Code on Pro, Max, Team and Enterprise plans.

What Anthropic actually shipped

Opus 5.5 takes over from Opus 5 as the model Anthropic tells developers to start with for most work. The launch post is unusually frank about benchmarks: Anthropic writes that at this level, benchmark margins have become a weaker guide to real differences, and that in its own use the gap to Fable 5.1 is smaller than the scores suggest.

So the headline is efficiency, not a leap in raw ability. The model needs less computing power to serve, uses fewer tokens per task, and Anthropic passed part of that on in its pricing.

A cache read is text the model has already seen in the same job, such as your project files, which it rereads at a discount. In long agent sessions that is most of the bill, which is why the 60% cut matters more than the headline price. There is also a fast mode, which Anthropic says runs up to 2.5 times faster, at $8 and $40 per million tokens in Claude Code and on the Claude Platform.

What changes if you are not a developer

You do not need an API key to notice this release. Three changes reach ordinary Claude subscribers:

• More room before you hit a limit. Anthropic raised the five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans. Subscribers also got a rate limit reset they can save and spend whenever they choose, for example on a deadline day.
• Answers you can scan. Early testers called the writing clearer and easier to follow. It leads with the most important point and uses less jargon, which matters when its work goes straight to a client or a manager.
• Fewer made-up numbers. In one internal test, Claude had to report on a company's quarterly results from a copy of the web where the earnings release was hard to find, and a grader failed any report with an invented figure or quote. 16 of 18 Opus 5.5 reports passed. Neither Fable 5.1 nor Opus 5 passed in any attempt.

That last result is one internal test, not a promise. Keep checking every number before it goes into a proposal, a price list or an invoice.

Where Opus 5.5 sits in the new lineup

Within 15 days the rest of the family arrived, so the model menu looks different from a month ago.

Independent boards back up the launch claims so far. In our AI model leaderboard snapshot, Opus 5.5 is first on LMArena's Code Arena, where people vote on which model built the better web page, and it shares first place on the main text board. It has far fewer votes than older models, so those ranks can still move.

How to try it this week

If you use Claude Code on a paid plan, you are probably on Opus 5.5 already: the Claude Code model settings list it as the default for Pro, Max, Team, Enterprise and API accounts. Type /model to check which model is active, and /model opus to switch back if you changed it.

New to the tool? Start with our plain-English guide to what Claude Code is, then follow the install walkthrough.

Then test the two claims that matter most, clearer writing and fewer invented figures, on a real file from your own week:

Read [THE REPORT, SPREADSHEET OR EXPORT YOU ATTACHED]. Write a one-page summary for [WHO WILL READ IT].
Put the single most important finding in the first sentence.
After every number, say exactly where it came from (page, tab or cell).
If something is missing or unclear, list it under "Open questions" instead of guessing.

Compare the result with what the same file produced last month. Look at two things: how fast you find the main point, and whether every number traces back to your file.

What to watch next

• The price race. Google's Gemini 4 Argon doubles its API price on December 1, 2026, and OpenAI's GPT-6.1 Sol starts at the same $2 and $10 per million tokens as Sonnet 5.5. Expect more price moves before the end of the year.
• Arena ranks. As votes pile up on LMArena, Opus 5.5's three-way tie at the top of the text board will either hold or break.
• Safeguard friction. If your work touches security testing or biology research, watch whether the rerouting to older models slows you down, and whether Anthropic widens its verification programs as promised.