September 28, 2026, brought Claude Sonnet 5.5 to Anthropic’s Claude lineup, including Claude.ai on the web. Anthropic presents Sonnet as a faster, lower-cost choice for everyday work alongside its more demanding Opus model. If you use Claude to untangle a long document or get a draft moving, the interesting question is simple: Will you notice the upgrade?
What Happened
Anthropic launched Sonnet 5.5 as an upgrade over Sonnet 5. The company says the new model generates output more than 30% faster and can cost up to 30% less per task for typical work. The base API rates remain the same as Sonnet 5; Anthropic attributes the task savings to using fewer tokens. Those are Anthropic’s claims, and “typical” matters. Your result depends on the task and how much the model writes back.
The company points to writing, analysis, documents, spreadsheets, and presentations as useful jobs for Sonnet. That makes this a relevant launch even if you’ve never touched an AI developer tool. Waiting for a summary or asking Claude to revise a presentation can feel like watching a progress bar crawl across the screen. A faster response would be welcome (assuming the answer is still good).
Anthropic’s public Sonnet page is the official place to see how it presents the model. Its news page and API release notes are useful for checking subsequent announcements.
Who Can Use It?
Sonnet 5.5 is available through Claude.ai. Check the model choices and subscription terms shown in your account before changing plans; the API rates below are a separate form of billing.
For developers and businesses using the API, the listed Sonnet 5.5 rates are $2 per million input tokens, $10 per million output tokens, and $0.20 per million cache-read tokens. Tokens are pieces of text: what you send counts as input, and what Claude writes counts as output. These are API rates, not a Claude.ai subscription price. Check Anthropic’s API documentation for current terms before estimating a bill.
How It Stacks Up
The clearest comparison with Sonnet 5 is speed. Anthropic says Sonnet 5.5 produces output more than 30% faster. That could be noticeable when Claude has a lot to write, such as a long summary or a draft with several sections. A quick, one-sentence answer may give you little reason to notice. Anthropic also describes Sonnet 5.5 as more efficient on typical tasks, but that is not a promise that every request costs less.
The comparison with Opus 5.5 is more delicate. Anthropic’s launch comparison reports 1,844 for Sonnet 5.5 and 1,846 for Opus 5.5 on GDPval-AA v2.1, an occupational-work evaluation. These are benchmark scores, not percentages or a general ranking of model quality. That’s a tiny gap on that test. Think of it like two cars posting nearly identical times on one track: useful information, but it doesn’t tell you how they’ll handle every road. Anthropic still positions Opus for more demanding, open-ended work.
| Model | What the available evidence says | What it means for everyday use |
|---|---|---|
| Claude Sonnet 5 | The preceding Sonnet model; Anthropic says Sonnet 5.5 generates output more than 30% faster. | A reason to try the newer model when you’re waiting on longer answers. |
| Claude Sonnet 5.5 | Scored 1,844 on the GDPval-AA v2.1 occupational-work benchmark. Anthropic lists API rates of $2 per million input tokens and $10 per million output tokens in its API information. | A promising choice for routine writing, summaries, and document work. |
| Claude Opus 5.5 | Scored 1,846 on that same benchmark. API rates are $4 per million input tokens and $20 per million output tokens; Anthropic positions Opus for more demanding work. | Keep it in mind when a task is complex enough that you want to compare both answers. |
The per-token rate is only part of the bill. A longer answer uses more output tokens, so a model’s efficiency claim does not guarantee savings on every request. Compare the cost of a finished task as well as the price per million tokens.
This comparison focuses on Sonnet and Opus. One occupational-work benchmark does not establish equal performance across coding, reasoning, factual accuracy, or every consumer task, and it cannot by itself settle a ranking against rival models.
Our Take
Verdict: If you regularly use Claude for drafts, summaries, or spreadsheet and presentation work, Sonnet 5.5 is worth trying when it appears in your account. Anthropic’s speed claim gives you a good reason to compare it with Sonnet 5. The near-Opus benchmark gives you a reason to test it on harder work, but not a reason to assume the two models are equal.
How to Check Availability
On Windows or Mac, open Claude.ai in your browser and check the model choices offered in your account. If Sonnet 5.5 is available, try a familiar document or task, review the facts in the answer, and compare the complete result as well as the waiting time.
Sonnet 5.5 looks most useful to people who spend a lot of time waiting for Claude to draft or summarize. My take: try it for routine work if your account offers it, and compare it with Opus on tasks where the answer really matters. The close benchmark score is a good reason to be curious; your own results are the better reason to switch.