Claude Sonnet 5 is positioned by Anthropic as an everyday agent model aimed at coding, planning, browsing, and general productivity tasks. It sits in the claude-sonnet family and was released on June 30, 2026, with a knowledge cutoff in early 2026 and closed weights, meaning the model is only accessible through hosted providers rather than as a downloadable checkpoint. The catalog notes a regional EU distribution routed through European infrastructure, which is useful for teams with data-residency requirements in Europe.

The model's defining practical strength is a very large working memory paired with a generous response budget, allowing long documents, multi-file codebases, or extended agent loops to be handled in a single turn without aggressive truncation. Across the providers tracked on models.dev, tool calling and reasoning are consistently supported, while structured-output availability varies by host. This combination of long context, tool use, and reasoning makes the model well suited to agentic workflows such as repository-scale code edits, multi-step research and planning, and browser-driven task automation, especially when the task requires sustained coherence over many intermediate steps.