News in Short

• Anthropic has launched Claude Sonnet 5.5 as the successor to Sonnet 5.
• The company claims the new model is 30% faster than its predecessor.
• Anthropic says Sonnet 5.5 costs 30% less per task in internal testing.
• The model retains Sonnet 5’s token pricing.
• Claude Sonnet 5.5 brings improved coding and knowledge-work capabilities.
• It is available through AWS, Google Cloud, Microsoft Azure and Claude Platform.

Anthropic has launched Claude Sonnet 5.5 as the latest addition to its Claude model lineup. The company says the new model improves performance while reducing the cost of completing tasks. It also brings stronger coding capabilities and improvements across knowledge-work tasks.

Importantly, Anthropic has kept the model’s token pricing unchanged from Claude Sonnet 5.

Claude Sonnet 5.5 Claims 30% Faster Performance

Anthropic says Claude Sonnet 5.5 is 30% faster at generating outputs than Sonnet 5. The company describes it as its fastest model yet.

The model also aims to reduce the amount of computing required to complete tasks. According to Anthropic’s internal testing, Sonnet 5.5 costs 30% less per task than its predecessor.

This means developers pay the same token rates while potentially using fewer resources for comparable tasks.

Claude Sonnet 5.5 Price and Availability

Claude Sonnet 5.5 retains the same pricing as Sonnet 5.

It costs $2 per million input tokens and $10 per million output tokens. Cache reads cost $0.20 per million tokens, while cache writes cost $2.50 per million tokens.

Anthropic says the model is also cheaper than Claude Opus 5.5.

Developers can access Claude Sonnet 5.5 through the Claude Platform. It is also available through Amazon Web Services, Google Cloud and Microsoft Azure.

Coding Performance Gets a Boost

Coding is one of the areas where Anthropic says Sonnet 5.5 has made significant progress.

On Terminal-Bench 4.0’s Agentic Coding test, Claude Sonnet 5.5 scored 70.6%. Claude Sonnet 5 scored 10.3%, while Opus 5.5 scored 66.4%.

The model also scored 46.2% on FrontierCode 1.1. For comparison, Opus 5.5 scored 54.4% and OpenAI’s GPT-6 Sol scored 49.3% on the same test.

These figures come from Anthropic’s reported evaluations. Benchmark results can vary depending on testing methodology and model configurations.

Improved Knowledge Work and Cyber Capabilities

Beyond coding, Anthropic says Claude Sonnet 5.5 improves performance across several knowledge-work tasks.

The company also claims improvements in cyber capabilities compared with Sonnet 5. The model can detect and fix bugs in code as part of its coding capabilities.

Anthropic says Sonnet 5.5 improved or matched Sonnet 5 across most measures in its automated behavioural audit. These included alignment, resistance to misuse and honesty.

The company is deploying the model with the same safeguards used for Opus 5.5.

Same Pricing With More Efficient AI Workloads

The launch puts the focus on efficiency as much as raw model performance. Anthropic is positioning Sonnet 5.5 as a model that can complete more demanding work while using fewer resources.

The company claims the model can deliver outputs 30% faster. It also says each task costs 30% less internally compared with Sonnet 5.

For developers building coding tools, AI agents and other knowledge-work applications, these changes could affect both response times and operating costs.

However, the reported efficiency figures come from Anthropic’s internal testing. Actual costs and performance can vary based on workload, usage patterns and deployment environment.