Claude Sonnet 4 (latest)

Claude Sonnet 4 is a hybrid reasoning model designed to deliver fast, capable intelligence for real-time agents and high-volume workloads. It represents a significant step forward from its Sonnet 3.7 predecessor, balancing frontier-level capability with computational efficiency rather than pursuing raw power alone. The model is optimized for practical everyday use cases, from routine coding tasks to complex software development projects, excelling at autonomous codebase navigation and following intricate instructions with improved precision and controllability. Its state-of-the-art performance on SWE-bench at 72.7% demonstrates competitive coding ability, while reduced error rates in agent-driven workflows make it a reliable choice for developers building automated systems.

Building on the foundation established by Sonnet 3.7, Sonnet 4 introduces extended thinking with tool use, allowing the model to alternate between reasoning and tool invocation during complex tasks. The model gained the ability to use tools in parallel and demonstrated significantly improved memory capabilities, extracting and saving key facts to maintain continuity across extended sessions. A major capability leap came with support for up to 1 million tokens of context—a fivefold increase enabling developers to process entire codebases exceeding 75,000 lines or analyze dozens of research papers without losing track of relationships and dependencies. This expanded context window opens doors to large-scale code analysis and demanding applications that were previously impractical, positioning Sonnet 4 as a practical workhorse for developers and researchers handling complex, long-running projects.