Claude Sonnet 5.5 Operational

Claude Sonnet 5.5 is the second model in Anthropic's Claude 5.5 series, released just 6 days after the flagship Claude Opus 5.5. It is a comprehensive upgrade over Claude Sonnet 5, positioned as a complementary and more affordable alternative to Opus 5.5 for clearly defined everyday tasks. Claude Sonnet 5.5 achieves 70.6% on Terminal-Bench 4.0, approaching the flagship on most benchmarks. It excels at everyday programming, bug fixing, and knowledge work such as documents, presentations, and spreadsheets.

High-performance coding: Achieves 70.6% on Terminal-Bench 4.0, supports multi-step professional command-line tasks, and surpasses Opus 5.5.
Everyday knowledge work: Excels at clearly defined tasks such as bug fixing, document creation, presentations, and spreadsheets, while also demonstrating strong design sensibility.
Codebase comprehension: Quickly understands large codebases and can process tens of thousands of lines of game architecture code reliably over extended periods.
Cost-effective operation: Priced the same as Sonnet 5, it can exceed the previous generation's best results even at Low or Medium effort, with per-task costs of about one-tenth of the previous generation.
Enterprise-grade agent applications: Can be embedded in scenarios such as Rovo Agent, Slackbot, and customer service systems to reduce decision errors and improve ticket-handling and operational efficiency.
Vision and game understanding: The first Sonnet model to complete Pokémon Red using only screenshots, with significantly enhanced multimodal capabilities.
Secure cybersecurity handling: The first Sonnet equipped with cybersecurity guardrails; high-risk cybersecurity tasks automatically fall back to Sonnet 5.

Effort control mechanism: The model supports multiple reasoning levels—Low, Medium, High, Xhigh, and Max—balancing cost, speed, and quality by controlling reasoning depth. Lower levels suit frequent everyday tasks, while higher levels support complex reviews and long-chain decisions, allowing the same model to cover multiple scenarios.
Parallel tool calls and batching: Sonnet 5.5 combines tool calls that can run in parallel into batches, reducing serial wait times. In testing, tool calls decreased by about one-third and shell executions nearly halved, directly reducing latency and token consumption.
Sub-agent task decomposition: At higher reasoning levels, the model invokes Claude Code's code review Skill and delegates the review to multiple sub-agents for parallel processing.
Cost-efficiency architecture: The model reduces actual costs by decreasing the number of tokens required to complete tasks. According to official data, costs for most tasks can fall by up to another 30%, enabling more tasks to be completed within the same budget.
Safety routing and model fallback: An integrated cybersecurity assessment layer automatically routes high-risk security tasks back to Sonnet 5, maintaining everyday development capabilities while isolating abuse risks.

Access points: Use it directly through the Claude website/app, Claude Code, or Claude Platform. The API is also available on AWS, Google Cloud, and Microsoft Azure.
Switch models: In the conversation interface or model selector, switch from Sonnet 5 to Sonnet 5.5. The API model name is claude-sonnet-5-5.
Set the effort level: Choose Low, Medium, High, Xhigh, or Max based on task complexity. Medium is sufficient for everyday tasks and can save money and time.
Integrate with coding tools: Call it directly from tools such as Claude Code and Cursor, allowing it to process tool calls in parallel batches for greater efficiency.
API calls: Developers can call the API using claude-sonnet-5-5. Pricing is $2/M input tokens and $10/M output tokens, with zero data retention supported.

Near-flagship performance: Achieves 70.6% on Terminal-Bench 4.0, surpassing Opus 5.5's 66.4% and bringing mid-tier performance close to the flagship.
Exceptional cost-performance ratio: Its price is half that of Opus 5.5, while per-task costs can be as low as one-tenth of the previous generation.
Significantly faster: More than 30% faster than Sonnet 5, using fewer tokens for the same work and reducing actual costs by up to another 30%.
More efficient tool use: Processes tool calls in parallel batches, reducing tool calls by about one-third and nearly halving shell executions, with fewer steps and less risk of getting stuck.
Strong at both coding and knowledge work: Code changes can be merged directly, large codebases are understood quickly, and presentations, documents, and spreadsheets approach Opus-level quality.
Enterprise-grade reliability: Offers the lowest tool-call failure rate and remains responsive during long-running tasks, making it suitable for large-scale production deployment.

Official website: https://www.anthropic.com/claude-sonnet-5-5

Everyday programming and bug fixing: Handle clearly defined coding tasks in Claude Code and Cursor, with code changes ready to merge directly, fewer steps, and lower costs.
Large-codebase auditing: Quickly understand tens of thousands of lines of code and audit system architecture and data flows.
Enterprise knowledge work: Generate and iterate on documents, presentations, and spreadsheets in batches, producing operational review deliverables that are nearly ready to use and follow specified templates.
Customer service and business process automation: Embed it in systems such as Zendesk, Slack, and Atlassian Rovo to speed up ticket handling by about 20% and reduce erroneous decisions.