Anthropic debuts Claude Sonnet 4.5, leading coding benchmarks and offering a new SDK for building complex, long-running AI agents.
Anthropic has released Claude Sonnet 4.5, its latest LLM focused on coding and agentic tasks. The model sets new records on benchmarks like SWE-bench Verified and OSWorld, demonstrating superior performance in coding and general computer-use scenarios. A standout feature is its ability to maintain focus on complex, multi-step tasks for over 30 hours, a significant leap in persistent reasoning for AI systems. Alongside the model, Anthropic is offering the Claude Agent SDK, enabling developers to build custom agents with the same infrastructure that powers Claude’s own coding capabilities[5].
The release signals Anthropic’s commitment to advancing both the raw capabilities of LLMs and the tools available for developers to create sophisticated, autonomous agents. Improvements in reasoning and mathematical ability are also highlighted, broadening the potential applications in technical and enterprise environments. By open-sourcing key components of its agent infrastructure, Anthropic is lowering the barrier to entry for advanced AI agent development, while continuing to push the boundaries of what LLMs can achieve in complex, real-world workflows[5].
Source: https://www.edtechinnovationhub.com/news/anthropic-unveils-claude-sonnet-45-with-coding-agent-sdk-and-app-upgrades EdTech Innovation Hub