The llm-anthropic plugin has released version 0.25, introducing Claude Opus 4.7 as a new model option. This latest iteration supports a new "thinking_effort: xhigh" parameter, enabling users to leverage enhanced reasoning capabilities. The release also increases the default maximum token limit to align with the model's maximum capacity, allowing for longer and more comprehensive responses.
Two new boolean configuration options have been added: thinking_display and thinking_adaptive. These features give users greater control over how the model's internal reasoning processes are presented and adapted during interactions. The thinking_display option currently supports summarized output exclusively through JSON output formats or JSON logs, limiting its availability to these specific interfaces for now.
These updates represent a refinement of the llm-anthropic tool's functionality, offering developers and users more granular control over Claude's reasoning and output presentation. The enhanced thinking capabilities and increased token limits could enable more complex problem-solving tasks and longer-form content generation, making the plugin more versatile for advanced use cases requiring sophisticated AI reasoning.
Key Takeaways
- The llm-anthropic plugin has released version 0.
- This latest iteration supports a new "thinking_effort: xhigh" parameter, enabling users to leverage enhanced reasoning capabilities.
- The release also increases the default maximum token limit to align with the model's maximum capacity, allowing for longer and more comprehensive responses.
- Two new boolean configuration options have been added: thinking_display and thinking_adaptive.
Read the full article on Simon Willison
Read on Simon Willison