Claude Sonnet 4.6 beats Opus in agentic tasks, adds 1 million context, and excels in finance and automation, all at one-fifth ...
Identifying vulnerabilities is good for public safety, industry, and the scientists making these models.