Anthropic PBC today announced the launch of Claude Sonnet 5.5, the most capable mid-tier model in the company’s AI family, designed for everyday tasks and a clear upgrade over the previous generation, running over 30% faster at a lower cost.
Sonnet operates as the workhorse of Anthropic’s Claude family of models for everyday use and Sonnet 5.5 is well trained for tasks such as document generation, summarization, spreadsheets and fixing coding bugs. The company added that it also has a “strong eye for design,” allowing it to add polish to user interfaces and it can follow templates to create slides that need minimal editing.
This release follows the launch of Opus 5.5, the flagship model in the family, and incorporates clearer writing styles.
Although the company said the new Sonnet 5.5 writes more clearly and in a more human-like way, it also includes invisible text watermarking. When the company announced the new watermarking technology, it said it would not affect text quality or overall readability, but would increase the likelihood that AI-generated text could be detected. Anthropic added this capability to comply with global regulations, including the European Union’s AI Act.
Like Opus 5.5, Sonnet 5.5’s writing and communication clarity make it a better partner for collaboration. Anthropic said early testers described it as faster on iteration for less complex tasks and smarter than the previous generation.
The model is priced the same as Sonnet 5 at $2 per million input tokens, $10 per million output tokens, and $0.20 per million cache reads. According to the company, the model would typically require far fewer tokens to do the same work, making it 30% cheaper than the previous generation.
“We fed Claude Sonnet 5.5 hundreds of real support use cases across replies and escalation requests,” said Zendesk Inc. Director of AI Abhinay Kathuria. “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.”
On performance, the company said the new model scored 70.6% on Terminal-Bench 4.0, an agentic coding evaluation, compared to the previous model’s 10.3%. It also scored two points below Opus 5.5 on GDPval-AA, a test of real-world occupational work and tasks. Anthropic added that this was the first Sonnet model to beat Pokémon Red, a video game commonly used to test long-horizon work and image understanding, using only screenshots.
Anthropic noted that Sonnet 5.5 is more than 30% faster than the previous generation in output speed, making it the fastest Sonnet model to date. The company added that it plans to release Haiku 5.5, the smallest, fastest and most efficient model in the Claude family built for high-volume and cost-sensitive applications, in the coming weeks.
Because the model is built on the same foundation as Opus 5.5, it is also the first Sonnet with notable cybersecurity capabilities. As a result, the company said it is the first Sonnet model to launch with cyber safeguards and fallbacks like those developed for more capable models such as Mythos-class models and Fable. These may trigger when prompts or questions touch on cybersecurity, biology, or other sensitive content, but most software development and life sciences work will be unaffected.