Claude 4.5 Sonnet is a frontier language model developed by Anthropic, released on September 29, 2025. It serves as a significant update to the Sonnet product line, optimized for high-volume production workloads that require a balance of speed, cost, and advanced intelligence. This version represents the standard, non-reasoning configuration of the model, designed for direct and efficient performance without the latency associated with extended thinking modes.
The model is specifically engineered for agentic workflows and complex autonomous tasks. Anthropic reports that the model can maintain focus for over 30 hours on multi-step software engineering and research operations. It achieved record-breaking performance on the SWE-bench Verified coding evaluation (77.2%) and leads the OSWorld benchmark for open-ended computer navigation (61.4%), demonstrating a high level of proficiency in tool orchestration and computer use capabilities.
Technically, Claude 4.5 Sonnet features a standard 200,000-token context window, with public beta support for up to 1 million tokens. It is a multimodal model capable of processing text, code, and image inputs. New system features include speculative parallel tool execution, enhanced memory management for long-horizon tasks, and a refined communication style that is more concise and direct compared to previous generations.
While the model is highly capable in reasoning and mathematics, this non-reasoning variant is tuned for predictable, low-latency responses. It is intended for use cases such as real-time customer support agents, large-scale code refactoring, and automated financial analysis where rapid tool interaction and high instruction-following accuracy are prioritized over visible chain-of-thought processing.