Choosing a Claude model and effort level in Claude Code (claude.com)

Claude Code has introduced a new model selection and effort level mechanism, allowing users to fine-tune how the AI handles tasks. Model selection determines which set of fixed weights, representing the model’s capabilities, is utilized, while effort level governs the amount of work Claude undertakes, including the number of files read and the complexity of actions performed. This enables users to choose smaller models for routine tasks and leverage larger models for more complex or ambiguous challenges. Default effort levels are recommended to optimize performance without overexerting resources, while adjustments should be made based on task complexity. Significantly, the ability to adjust both the model and effort level enhances the efficiency and accuracy of Claude Code in various programming scenarios. The model processes input by predicting tokens through computations involving billions of parameters, allowing it to assess context and actions taken. Claude's design now includes the capability to revise its action plan based on real-time feedback, promoting a more adaptive problem-solving approach. This flexibility is crucial for developers, as it assists in managing costs effectively while maintaining the quality of outputs—particularly for multi-step and demanding tasks, where larger models outperform their smaller counterparts.