Anthropic, a company specializing in artificial intelligence, has released Claude Sonnet 5.5, a new AI model that is faster and less expensive than its predecessor, Sonnet 5. According to the company, Sonnet 5.5 generates responses more than 30% faster than Sonnet 5 and costs up to 30% less per task. It has become the default model for the free version of Claude. The company explains that while it is slowing down the development of AI, it continues to release new versions. Less than a week after launching Opus 5.5, Anthropic introduced Sonnet 5.5, the second model in the 5.5 family. It follows Sonnet 5, released at the end of June, and will be succeeded by Haiku 5.5, which is expected to be announced in the coming weeks. Anthropic presents Sonnet 5.5 as a faster and more cost-effective addition to Opus 5.5, which is reserved for complex tasks. The company claims that its new model is "the most efficient for well-defined daily tasks," such as fixing software bugs or creating documents, presentations, and spreadsheets. It also emphasizes a refined design, resulting in more polished interfaces and presentations that follow provided templates. Early testers have also noted that the writing style of Sonnet 5.5 is clearer than that of Sonnet 5, according to the company. In terms of performance, Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agent coding test, compared to 10.3% for Sonnet 5 and 66.4% for Opus 5.5, which is more expensive. It is the first Sonnet model to include "frontier" type security measures as well as classifiers that prevent the extraction of its reasoning. On the GDPval-AA test, which evaluates real tasks across 44 professions, it scores 1844 points, compared to 1846 for Opus 5.5 and 1449 for Sonnet 5. The pricing for Claude Sonnet 5.5 on the API is $2 per million input tokens and $10 per million output tokens, which is half the cost of Claude Opus 5.5, priced at $4 for input and $20 for output. The default effort level varies depending on the environment: it is set to "medium" in applications and in Claude Code, and to "high" on the API. Developers transitioning from Sonnet 5 must also deal with five changes that could break their integrations, detailed in a technical blog post from Anthropic. The reflection feature, which is active by default, can no longer be completely disabled. It is now necessary to use the between_tools setting, which limits it to intervals between two tool calls. Sonnet 5.5 is the first Sonnet model to include cybersecurity safeguards previously reserved for Anthropic's most powerful models. The company states that its cybersecurity capabilities are comparable to those of Opus 5.5. In practice, bug fixing in a classical development framework is not affected, but the most sensitive cybersecurity requests are now more likely to be handled by Sonnet 5.5. The model finally includes classifiers that block the extraction of its reasoning, targeting distillation attacks, which involve extracting the capabilities of a large-scale model from thousands of fake accounts. Claude Sonnet 5.5 is already available: In Claude applications, where it replaces Sonnet 5 as the default model for the free offering, via the API, under the identifier claude-sonnet-5-5, on Amazon Bedrock, Google Cloud, and Microsoft Foundry, and in Claude Code, with the command /model sonnet (Opus 5.5 remains the default model). The pricing for Claude Sonnet 5.5 on the API is $2 per million input tokens, $10 per million output tokens, and $0.20 for reading a million tokens in cache. Anthropic acknowledges, however, that these scores "reflect only one aspect of a model's capabilities," and that Opus 5.5 remains significantly superior for complex and open-ended work that requires prolonged judgment. The model is intended to be more efficient, performing the requested tasks with fewer tokens and generating responses more than 30% faster. Sonnet 5.5 is immediately available on all platforms, including the API, without data retention.