Claude Sonnet 5.5: Same $2/$10 Price, Fewer Tokens
Claude Sonnet 5.5 launched on 28 September at unchanged $2 and $10 prices per million tokens. What changed, what to test, and where Opus 5.5 still wins.
Anthropic released Claude Sonnet 5.5 on Monday 28 September 2026, and the list price did not move: $2 per million input tokens and $10 per million output tokens, the same as Sonnet 5. The change is in how much work a dollar buys. Anthropic says the model generates output more than 30% faster and costs up to 30% less per task, because it needs fewer tokens and fewer tool calls to get to the same result. The model ID is claude-sonnet-5-5.
What Claude Sonnet 5.5 costs
The price sheet, as reported by Unite.AI:
Input: $2 per million tokens
Output: $10 per million tokens
Cache reads: $0.20 per million tokens
Cache writes: $2.50 per million tokens
Nothing on that list is lower than before. If your usage is identical to last week, your bill is identical to last week. Any saving comes from the model doing the same job in fewer steps, which is the same distinction we broke down for Opus in three price cuts and two catches and in our explainer on token efficiency.
Where it sits next to Opus 5.5
Gizmodo describes it as a faster, cheaper alternative to Opus 5.5, which debuted less than a week earlier. Anthropic positions Sonnet 5.5 for well-scoped everyday work: fixing bugs, and producing polished documents, slides and spreadsheets. Opus 5.5 stays the pick for complex work that needs careful judgment.
That split is a useful default, not a rule. A well-scoped bug fix on a clean codebase is a Sonnet job. A vague brief that needs the model to decide what the real problem is still leans toward the bigger model. If you route between tiers already, this is a reason to re-check where your line sits, and our guide on which AI model for which task covers how to draw it.
The benchmark numbers, and one to be careful with
Anthropic's reported figures, again via Unite.AI:
Terminal-Bench 4.0: 70.6%, against 10.3% for Sonnet 5
CursorBench 4.0: 55.5%
Humanity's Last Exam: 64.5% with tools
The Terminal-Bench line is the striking one. A jump from 10.3% to 70.6% between two point releases is far larger than the gaps between the rest of the numbers, and a jump that size can come from the model, from a change in how the test was run, or from both. Anthropic's own Opus 5.5 launch a week earlier put that model at 66.4% on the same benchmark, so the new Sonnet landing above it is worth a second look before it drives a purchasing decision. We wrote up how to tell when a benchmark score is misleading for exactly this situation.
Safety and availability
Sonnet 5.5 is the first Sonnet model to ship with the cyber safeguards and fallbacks used on Anthropic's most capable models, with a reported 99.43% recall on the company's cyber harm coverage set. It is available on Amazon Web Services, Google Cloud and Microsoft Azure, and is offered with zero data retention.
Anthropic also said Claude Haiku 5.5 will join the 5.5 family in the coming weeks. If your workload is high-volume and simple, that is the release to wait for before re-pricing.
What to do this week
Do not switch on the strength of a launch post. Take 20 real prompts from your own logs and run them through both Sonnet 5 and Sonnet 5.5.
Compare tokens used and tool calls per task, not just answer quality. That is where the claimed 30% lives.
If you have pinned a model version, read whether you should pin an AI model version before moving the pin.
Keep the result. Vendors will keep shipping point releases every few days at this pace, and a saved test set turns each one into a ten-minute check. More on keeping pace without burning out in how to keep up with AI news.
FAQ
Is Claude Sonnet 5.5 cheaper than Sonnet 5?
Per token, no. The list prices are identical. Anthropic says it is up to 30% cheaper per task because it uses fewer tokens and tool calls.
What is the Claude Sonnet 5.5 model ID?
The API model ID is claude-sonnet-5-5.
Should I use Sonnet 5.5 or Opus 5.5?
Use Sonnet 5.5 for well-scoped, repeatable work where speed and cost matter. Use Opus 5.5 for complex tasks that need careful judgment. Test both on your own prompts before committing.
When does Claude Haiku 5.5 arrive?
Anthropic has only said it is coming in the coming weeks, with no date given.
How did this land?
About the author
Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.
