Anthropic's Claude Sonnet 4.5 is available now - ‘the best AI model in the world for real-world agents, coding, and computer use’ (www.techradar.com)

Anthropic today launched Claude Sonnet 4.5, a generational update the company touts as its “best coding model” and “strongest model for building complex agents and using computers.” The release is accompanied by feature upgrades across Anthropic’s products: checkpoints for Claude Code (save/roll back progress), in-chat code execution and direct file creation for spreadsheets, slides and documents, and wider availability of Claude for Google Chrome to waitlisted users. Pricing remains unchanged per million tokens and these new capabilities will be available on paid plans. Technically, Anthropic emphasizes improved alignment and long-horizon autonomy—claiming Sonnet 4.5 substantially reduces behaviors like sycophancy, deception, power-seeking and delusional suggestions—and offers much longer autonomous operation (about 30 hours vs. Opus 4’s 7 hours). The company reports benchmark and real-world task gains over Sonnet 4.1 (which had earlier bested several competing models), and highlights natural‑language agent construction, better handling of very long contexts, and workflow integrations that let models create, execute and manage code and files with persistence. For AI/ML teams this signals stronger off-the-shelf tooling for large-scale agentic workflows, more reliable coding assistants, and practical guardrails for safer, longer-running autonomous systems.