Anthropic Releases Claude Sonnet 5.5 With 30% Speed Boost and Lower Pricing

Anthropic PBC introduced Claude Sonnet 5.5 as its most capable mid-tier offering, positioning the model as a significant step forward from the previous generation with performance gains exceeding 30% and reduced operational expenses.

Within Anthropic's Claude lineup, Sonnet serves as the primary workhorse for routine applications. Sonnet 5.5 excels at document generation, summarization, spreadsheet work and code debugging. The model also demonstrates strong visual design capabilities, enabling it to enhance user interface elements and generate presentation slides requiring minimal revision.

The release arrives following the September 22 introduction of Opus 5.5, Anthropic's flagship model. Sonnet 5.5 features enhanced writing quality and a more natural communication style.

While the new model produces clearer and more human-like text, it incorporates invisible text watermarking technology. Anthropic stated that this watermarking would not compromise text quality or readability but would improve the detectability of AI-generated content. The company implemented this feature to align with international regulatory requirements, particularly the European Union's AI Act.

Like Opus 5.5, Sonnet 5.5's improved writing and communication capabilities position it as a stronger collaborative tool. Early users reported that the model delivers faster iteration on simpler tasks and outperforms its predecessor in capability.

Pricing remains unchanged from Sonnet 5 at $2 per million input tokens, $10 per million output tokens, and 20 cents per million cache reads. Anthropic notes that the model typically consumes substantially fewer tokens to accomplish equivalent work, resulting in a 30% cost reduction relative to the previous generation.

We fed Claude Sonnet 5.5 hundreds of real support use cases across replies and escalation requests. It made fewer wrong decisions and resolved tickets faster than the Claude models we use in production today. Tickets were processed 20% faster, getting our customers the help they need without the wait.

Abhinay Kathuria, Director of AI at Zendesk Inc.

On standardized benchmarks, Sonnet 5.5 achieved 70.6% on Terminal-Bench 4.0, an agentic coding evaluation, compared to 10.3% for the prior version. The model scored two points below Opus 5.5 on GDPval-AA, which measures real-world occupational performance. Notably, Sonnet 5.5 became the first Sonnet model to successfully complete Pokémon Red, a video game frequently used to evaluate long-horizon reasoning and image comprehension, relying solely on screenshots.

Output speed improvements exceed 30% compared to the previous generation, making Sonnet 5.5 the fastest Sonnet model released to date. Anthropic indicated that Haiku 5.5, the smallest and most cost-efficient model in the Claude family designed for high-volume, price-sensitive use cases, will launch in the coming weeks.

Built on the same architecture as Opus 5.5, Sonnet 5.5 marks the first Sonnet model with substantial cybersecurity capabilities. Consequently, it launches with protective measures and fallback systems comparable to those implemented in more advanced models such as Mythos-class and Fable variants. These safeguards activate when prompts involve cybersecurity, biology or other sensitive domains, though most software development and life sciences applications will operate without restriction.