Anthropic Releases New Model "Claude Sonnet 5"

Anthropic has released a new AI model "Claude Sonnet 5". Positioned in the company's mid-tier segment, this model offers near-flagship performance while being available at significantly lower prices. In both pricing structure and performance metrics, it narrows the gap with higher-tier models.

As competition in the AI model space intensifies, companies are intensifying efforts to simultaneously achieve "high performance" and "low cost". Anthropic itself is entering a period of preparation for IPO (Initial Public Offering), aiming to expand its user base by making the model available to a broader range of developers and enterprises. The release of Sonnet 5 can be understood as part of this strategic direction.

On the pricing front, during the API adoption period (through August 31), the model is priced at $2 per million input tokens and $10 per million output tokens. After this period ends, pricing will transition to $3 for input and $15 for output, which still falls well below the top-tier model "Opus 4.8" at $5 for input and $25 for output. It will be the default model for Free and Pro plans, while also being available on Max, Team, and Enterprise plans.

On the performance front, multiple evaluation metrics show Sonnet 5 achieving significant improvements over its predecessor "Sonnet 4.6" and approaching Opus 4.8's capabilities. On "SWE-bench Pro", which measures AI coding ability, Sonnet 5 achieved 63.2% compared to Sonnet 4.6's 58.1%, narrowing the gap with Opus 4.8's 69.2%. On another coding evaluation, "Terminal-Bench 2.1", Sonnet 5 achieved 80.4%, nearly matching Opus 4.8's 82.7%.

On "Humanity's Last Exam", which measures reasoning ability across multiple domains, Sonnet 5 achieved 57.4% under the condition of tool use, nearly equivalent to Opus 4.8's 57.9%. Additionally, on the knowledge work benchmark "GDPval-AA v2", Sonnet 5 achieved a score of 1,618, slightly exceeding Opus 4.8's 1,615. These results demonstrate that at roughly 60% cheaper than standard pricing, the model reaches a performance range that overlaps with the flagship model.

What Anthropic particularly emphasized this time is the capability of "agentic AI". Agentic AI refers to a system that not only answers questions but can also autonomously execute multi-step tasks while using tools such as browsers and terminals. As of 2026, the role enterprises seek from AI is shifting from "conversation" to "autonomous task execution", and Sonnet 5 is positioned as a model that addresses this demand.

The release of Sonnet 5 can be seen as an attempt to broaden access to high-performance AI from "something only available to large enterprises" to "something accessible to cost-conscious developers". Considering Anthropic's business expansion and IPO preparation, future attention will focus on the expansion of user numbers and the extent to which agentic AI penetrates actual business operations.