Anthropic Releases Sonnet 5.5 AI Model at Half the Price of Opus

Artificial intelligence startup Anthropic released a new model called Sonnet 5.5 on Monday, marking the company's second deployment since Chief Executive Officer Dario Amodei publicly advocated for a slower pace of frontier AI development, according to a report from CNBC Business (https://www.cnbc.com/2026/09/28/anthropic-sonnet-5-5-launch.html).

Positioned as a faster and more economical option, Sonnet 5.5 is designed to outperform its predecessor in writing computer code, completing scoped tasks, and generating workplace documents, presentations, and spreadsheets. The launch comes less than a week after Anthropic debuted its higher-cost Opus 5.5 model, with the company adding that Haiku 5.5, its entry-level offering, will follow soon.

"Sonnet is really for the cost-conscious customer where they might not need as much intelligence," Theo Chu, a research product manager at Anthropic, told CNBC Business in an interview. "It might be routine tasks that just need execution, but don't need that judgment that Opus can bring."

The release comes amid broader industry scrutiny over catastrophic risks associated with cutting-edge AI. Earlier this month, Amodei called on AI companies to slow the pace at which they improve their most advanced systems.

Anthropic stated in a corporate blog post that Sonnet 5.5 does not push the frontier of overall model capability. Because of this, the company focused its alignment evaluations on a targeted set of baseline risks applicable across capability tiers, rather than frontier-level dangers.

However, Anthropic noted that Sonnet 5.5 features a "large improvement" in cybersecurity capabilities compared to its predecessor. Consequently, it is the first Sonnet-tier model equipped with the cyber safeguards and fallback mechanisms previously reserved for the company's flagship systems. Chu added that alignment and safety prioritization have remained central to the firm's mission across all tiers.

Sonnet 5.5 is priced at $2 per million input tokens and $10 per million output tokens—half the cost of Opus 5.5. The company stated the model also consumes fewer tokens per task than the previous version, reducing overall compute costs. The model launched Monday with immediate availability across Amazon Web Services, Google Cloud, and Microsoft Azure.