Claude Sonnet 5.5 Launch: Near-Opus Scores at Sonnet Prices
Claude Sonnet 5.5 launched 28 September 2026 at $2/$10 per million tokens, scoring close to Opus 5.5. What changed, the API caveats and who should switch.
PromptWises is reader-supported. Some links may earn us a commission at no extra cost to you. This never affects our ratings or recommendations.
Anthropic released Claude Sonnet 5.5 on 28 September 2026 as the mid-priced model of its 5.5 generation. It keeps Sonnet 5’s price of $2 per million input tokens and $10 per million output tokens, and by Anthropic’s own benchmarks it scores close to Opus 5.5, which costs twice as much. For most teams paying for Opus on everyday coding and agent work, Sonnet 5.5 is now the first model to test. Sources: Anthropic’s Sonnet 5.5 announcement and the model documentation.
What Anthropic launched
Sonnet 5.5 is the second model in the Claude 5.5 family, after Opus 5.5 on 22 September and before Claude Haiku 5.5 on 7 October. Anthropic’s release notes call it “a faster, lower-cost complement to Claude Opus 5.5”. The documentation gives these specifications:
- Context window: 1 million tokens; maximum output 128,000 tokens (up to 300,000 on the Batch API with a beta header).
- Inputs: text and images; output is text.
- Effort levels: Low, Medium, High, Xhigh and Max. The default is High on the API and Medium in the Claude apps and Claude Code.
- Model IDs:
claude-sonnet-5-5on the Claude API, Google Cloud, Microsoft Foundry and Claude Platform on AWS;anthropic.claude-sonnet-5-5on Amazon Bedrock. - Retirement: not sooner than 28 September 2027.
Anthropic also says Sonnet 5.5 generates output more than 30% faster than Sonnet 5 and usually needs fewer tokens for the same job, cutting cost per task by up to 30%.
Pricing
| Per million tokens | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Input | $2.00 | $2.00 | $4.00 |
| Output | $10.00 | $10.00 | $20.00 |
| Cache reads | $0.10 | $0.20 | $0.20 |
| Cache writes (5 min) | $2.50 | $2.50 | $5.00 |
List prices at the time of writing, from Anthropic’s pricing page. Sonnet 5.5 launched with $0.20 cache reads; Anthropic halved them to $0.10 on 7 October, which it estimates makes most agentic tasks on Sonnet 5.5 about 20% cheaper. The Batch API takes 50% off input and output.
How it performs, according to Anthropic
These are vendor-reported scores on tests Anthropic chose, so treat them as a reason to run your own evaluation rather than as proof. Our guide to reading AI model announcements explains why.
| Benchmark (Anthropic-reported) | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% | 66.4% |
| OSWorld 2.1 (subset) | 80.1% | 57.0% | 81.8% |
| CursorBench 4.0 | 55.5% | 34.1% | 57.8% |
| Humanity’s Last Exam (with tools) | 64.5% | 54.9% | 67.7% |
| GDPval-AA v2.1 (Elo) | 1844 | 1449 | 1846 |
Anthropic also reports that at Low or Medium effort Sonnet 5.5 beats Sonnet 5’s best score on several benchmarks at about a tenth of the cost per task. Opus 5.5 keeps a clear lead on FrontierCode, Anthropic’s harder coding test (54.4% against 46.2% at Max effort).
What breaks when you migrate
The model page lists five breaking changes for code written for Sonnet 5. Forced tool use now returns an error. Thinking blocks are tied to the model and conversation that produced them. The older computer_20251124 tool is rejected on the Claude API and Google Cloud. Setting temperature, top_p or top_k to a non-default value returns a 400 error. And if you run Sonnet with thinking off, Anthropic says to switch to the new between_tools setting, which turns off up-front thinking, before moving over. Read the migration guide before you change a production model ID.
Also new: API credits for Max and Team
Since 7 October, Max and Team subscribers get monthly credits for the Claude API, according to Anthropic’s help article. Max 5x includes $100 a month and Max 20x $200. Team plans get $20 per Standard seat and $100 per Premium seat, pooled and capped at $500 a month. You claim them by linking a Claude Console organisation in billing settings after seven days on the plan. Unused credits do not roll over. They cover the API, Managed Agents and the Agent SDK, but not interactive Claude Code or Claude on Bedrock, Vertex AI or Foundry. Free, Pro and Enterprise are not eligible.
Who should care
API builders on Sonnet 5 should test Sonnet 5.5 now: same per-token price, cheaper caching and Anthropic’s claim of fewer tokens per task. Teams paying Opus prices for routine coding should compare the two on their own tasks; the price gap is 2x. App users will mostly notice speed. Our Claude review covers what each plan includes, and our ChatGPT vs Claude comparison helps if you are still choosing. More launches are tracked in our news section.
Frequently asked questions
How much does Claude Sonnet 5.5 cost?
At the time of writing, $2 per million input tokens and $10 per million output tokens, the same as Sonnet 5. Cache reads are $0.10 per million tokens (cut from $0.20 on 7 October 2026), 5-minute cache writes $2.50, and the Batch API halves input and output prices.
Is Claude Sonnet 5.5 as good as Opus 5.5?
Close, by Anthropic's own numbers. It reports Sonnet 5.5 slightly behind Opus 5.5 on most of its benchmarks (for example 80.1% vs 81.8% on its OSWorld 2.1 subset) and ahead on Terminal-Bench 4.0. Opus 5.5 still leads on harder coding tests such as FrontierCode, and costs twice as much per token.
What is the Claude Sonnet 5.5 model ID?
claude-sonnet-5-5 on the Claude API, Google Cloud, Microsoft Foundry and Claude Platform on AWS, and anthropic.claude-sonnet-5-5 on Amazon Bedrock. It has a 1 million token context window and 128,000 tokens of maximum output.
Do I need to change my code to move from Sonnet 5 to Sonnet 5.5?
Possibly. Anthropic lists five breaking changes, including forced tool use returning an error and non-default temperature, top_p or top_k values returning a 400 error. If you run Sonnet 5 with thinking off, switch to the new between_tools setting first.
