Overview
Claude Sonnet 5.5 is Anthropic's mid-tier flagship model released on September 28, 2026, and the second product in the Claude 5.5 family after Opus 5.5. It is officially positioned for everyday, well-scoped work scenarios such as bug fixing, writing documentation, creating presentations and spreadsheets, with particular emphasis on sensitivity to design and visual layout. Compared with Sonnet 5, Sonnet 5.5 delivers over 30% faster output and up to 30% lower cost per task, while API pricing remains unchanged. This efficiency gain comes from completing tasks with fewer tokens, fewer tool calls, and fewer steps. The model ID is claude-sonnet-5-5, with a knowledge cutoff of June 2026, and it is simultaneously available on the Claude platform, Amazon Web Services, Google Cloud, and Microsoft Azure, with a zero data retention option.
Key Features
- Efficiency Gains and Cost Optimization: Output speed is over 30% faster than Sonnet 5, cost per task is up to 30% lower, while API pricing remains unchanged. The efficiency gain comes from completing tasks with fewer tokens, fewer tool calls, and fewer steps.
- Benchmark Performance: Terminal-Bench 4.0 score is 70.6%, compared with 10.3% for Sonnet 5 and 66.4% for Opus 5.5 at Xhigh effort; GDPval-AA v2.1 improved from 1449 for Sonnet 5 to 1844, close to Opus 5.5's 1846; FrontierCode 1.1 main set scores 46.2% at Max effort and 52.1% at Xhigh, compared with 42.4% for Sonnet 5; CursorBench 4.0 is 55.5%; OSWorld 2.1 partial score is 80.1%; Humanitys Last Exam with tools is 64.5%; Chartography without tools is 61.6%.
- Safety Protections: For the first time in the Sonnet series, near-top-tier cybersecurity protections are introduced, while bio-related protections remain at the Sonnet 5 standard.
- Platform and Integration: It is simultaneously available on the Claude platform, Amazon Web Services, Google Cloud, and Microsoft Azure, offers a zero data retention option, and Claude Code was integrated on the same day.
- Customer-Measured Results: Box reported a 2.4x speed improvement and 12% token savings; Slack's offline evaluation reduced output tokens by about 14%; Zendesk ticket handling sped up by 20%; Atlassian's Rovo Agents sped up by up to 30%; Balyasny Asset Management, across 2441 financial tasks, reduced tokens per answer from 497,000 for Sonnet 5 to about 121,000; Lovable reported about one-third fewer tool calls and roughly halved shell executions.
- Family Roadmap: Anthropic previewed that Haiku 5.5, the third model in the Claude 5.5 family, will be released in the coming weeks, focusing on high throughput and low cost.
Use Cases
- Bug fixing and code debugging
- Writing technical documentation and explanatory materials
- Creating presentations and spreadsheets
- Everyday tasks requiring sensitivity to design and visual layout
- Well-scoped automated workflows with relatively few steps
Pros
- Output speed is over 30% faster than Sonnet 5, cost per task is up to 30% lower, and API pricing remains unchanged
- Noticeable performance improvements on benchmarks such as Terminal-Bench 4.0, GDPval-AA v2.1, and FrontierCode 1.1
- For the first time in the Sonnet series, near-top-tier cybersecurity protections are introduced
- Simultaneously available on the Claude platform, Amazon Web Services, Google Cloud, and Microsoft Azure, with a zero data retention option
- Multiple customers' real-world tests show speed improvements and token savings, such as Box reporting a 2.4x speed improvement and 12% token savings
- Claude Code was integrated on the same day, making it easy for developers to use directly
Pricing
API pricing remains unchanged: $2 per million input tokens, $10 for output, $0.20 for cache reads, and $2.50 for cache writes. Other pricing information is subject to the official website.
Summary
Claude Sonnet 5.5 is Anthropic's mid-tier flagship model released on September 28, 2026, positioned for everyday, well-scoped work such as bug fixing, writing documentation, and creating presentations and spreadsheets. While keeping API pricing unchanged, it delivers over 30% faster output than Sonnet 5, up to 30% lower cost per task, and improved performance on multiple benchmarks. On safety, it introduces near-top-tier cybersecurity protections for the first time in the Sonnet series. For complex multi-step reasoning scenarios, the official recommendation is to use Opus 5.5.
Version History
- Cache read price halved (2026-10-08): Anthropic cut the Claude Sonnet 5.5 cache read price from $0.20 to $0.10 per million tokens. Since cache reads make up a large share of agentic work, the company estimates this lowers the cost of running most long-running tasks by about 20 percent
- Cursor announces Claude Haiku 5.5 pricing and lowers Claude Sonnet 5.5 cache read pricing. (2026-10-07): Cursor announced Claude Haiku 5.5 pricing at $0.10 per M input tokens and $0.50 per M output tokens, and $0.50/M and $2.50/M when input exceeds 100k tokens. Claude Sonnet 5.5 cache read pricing has also dropped from $0.20/M to $0.10/M, and users can compare Haiku 5.5's performance on CursorBench at cursor.com/evals.
- 用Claude Code和Amazon为受监管工作负载提供超强动力 (2026-10-05): Anthropic Claude Opus 5.5 and Claude Sonnet 5.5 are available on Amazon Bedrock in the AWS GovCloud (US) Regions. Learn how to use them with Claude Code, Anthropic's agentic coding tool, for complianc
- No. 3 on Agent Arena and No. 1 in Chat (2026-10-03): Arena reported Claude Sonnet 5.5 at No. 3 on Agent Arena with a +12.5% net improvement and a median cost of $2.74 per task, and No. 1 in the Chat category with +15.6%; Anthropic models took the top three spots on Agent Arena
- xHigh variant enters Code Arena and tops Coding Agent Index (2026-10-02): The xHigh variant of Claude Sonnet 5.5 entered the Code Arena: WebDev leaderboard at 1786, third place and 2 points behind GPT-6 Astra in second at 1788. Artificial Analysis also launched its Coding Agent Index, where Sonnet 5.5 (max) running in Claude Code leads with a score of 68 at 14.19 dollars per task, the highest cost tier on the index
- Code Arena ranking and limited Arena test (2026-09-30): Claude Sonnet 5.5 (High) ranked fourth on the Code Arena: WebDev leaderboard with 1699 points, while Arena opened a limited-time test with Direct Mode available for 48 hours
- Anthropic releases Claude Sonnet 5.5, 30 percent faster at unchanged pricing (2026-09-28): The second model in the Claude 5.5 family generates output over 30 percent faster than Sonnet 5 and cuts per-task cost by up to 30 percent, with API pricing unchanged at 2 dollars per million input and 10 dollars per million output tokens. It went live on AWS, Google Cloud and Azure at the same time, with Claude Code integrated the same day.