News Summary

Anthropic has unveiled Claude Sonnet 4.5, their latest AI model, which is being hailed as the safest and most advanced coding model in the world. This new model demonstrates significant improvements over its predecessors, with enhanced capabilities in coding, cybersecurity, finance, and autonomous tasks. Claude Sonnet 4.5 can autonomously work on projects for more than 30 hours, a notable enhancement from previous models. The model has undergone extensive safety training to reduce risks such as sycophancy, deception, and power-seeking behaviors, making it Anthropics’ safest system to date. It includes features like a refreshed terminal interface, file creation capabilities, and checkpoints for progress saving, aimed at improving user experience. API pricing for the model is set at $3 per million input tokens and $15 for the same amount of output tokens.