Anthropic Releases Claude Sonnet 5.5 AI Model

Anthropic has announced the release of Claude Sonnet 5.5, the second AI model in the Claude 5.5 lineup. The new release is positioned as a more affordable and faster solution for everyday tasks, programming, presentation generation, and document analysis.

How Fast Is It?

The main advantage of Sonnet 5.5 over Sonnet 5 is compute optimization. The model generates responses more than 30% faster and is up to 30% cheaper due to more efficient token usage for solving the same tasks.

At the same time, the pricing structure remains unchanged: $2 per 1 million input tokens and $10 per 1 million output tokens (0.20 per 1 million prompt-cached tokens). The model supports a context window of up to 1 million tokens and an output limit of up to 128k tokens.

Capabilities in Coding and Business Tasks

In programming benchmarks, the new model demonstrated massive growth:

• Terminal-Bench 4.0 (command-line task execution): 70.6% compared to just 10.3% for Sonnet 5.
• CursorBench 4.0: 55.5% versus 34.1% for the previous generation.
• FrontierCode 1.1 (at maximum effort): 46.2% for Sonnet 5.5 versus 42.4% for Sonnet 5.
• Although the flagship Opus 5.5 still leads in complex tasks requiring long reasoning chains, Sonnet 5.5 has come remarkably close to its level.

Sonnet 5.5 also introduces improvements at working with documents, spreadsheets, user interfaces, and presentations while strictly adhering to provided templates.

In real-world professional task evaluations (GDPval-AA v2.1), Sonnet 5.5 scored 1,844 points (compared to 1,449 for Sonnet 5), nearly matching the top-tier Opus 5.5 (1,846 points). In the AA-Briefcase benchmark, the result was 1,811 versus 1,359 points for its predecessor.

Vision and Computer Control

The model has improved its computer interface interaction capabilities:

• In the OSWorld 2.1 benchmark, its score rose from 57% to 80.1%.
• Recognition of charts and diagrams has improved.
• Sonnet 5.5 became the first model in the lineup capable of beating Pokémon Red relying solely on screen screenshots.

Safety and Protection

Due to increased capabilities in cybersecurity, Sonnet 5.5 received enhanced safeguards that were previously reserved for higher-tier models. For specific high-risk scenarios, the system can automatically fall back to Sonnet 5.

Additionally, the developers implemented anti-distillation classifiers that block high-volume requests designed to extract internal reasoning data to train third-party neural networks.

When and Where It Will Be Available

Claude Sonnet 5.5 is already available via the Claude web interface, API, as well as on Claude Platform, AWS, Google Cloud, and Microsoft Azure. 

In the coming weeks, Anthropic plans to launch Claude Haiku 5.5, the fastest and most budget-friendly model in the series designed for high-throughput workloads.