Simon Willison Updates LLM CLI with GPT-5.5 Integration and

Simon Willison has released version 0.31 of his popular LLM CLI tool, introducing native support for OpenAI's latest flagship model, GPT-5.5. This update, which arrived earlier this week, provides developers with a streamlined way to interact with the newest frontier models directly from the terminal while adding sophisticated controls over how those models respond.

The standout feature of this release is the llm 0.31 update, which enables the gpt-5.5 model flag for immediate use. Beyond simple model access, Willison has implemented a new verbosity system that allows users to dictate the length of model outputs. By using the -o verbosity option, developers can now toggle between low, medium, and high settings, a critical addition for those looking to reduce token waste or force more comprehensive explanations during complex debugging sessions.

Enhanced Vision and Custom Model Handling

Vision capabilities also receive a significant boost in the llm 0.31 update. The tool now supports an image_detail parameter, giving users the ability to specify the resolution of image attachments. Notably, for those utilizing GPT-5.5, a new "original" resolution setting is available, ensuring that high-fidelity visual data is preserved when the model analyzes technical diagrams or UI screenshots.

For power users who maintain their own model registries, the update standardizes asynchronous registration for custom models defined in extra-openai-models.yaml. This architectural shift improves the performance and reliability of the CLI when managing a diverse ecosystem of local and remote endpoints. The llm 0.31 update continues to solidify the tool's reputation as an essential utility for developers who prefer a keyboard-centric AI workflow.

This release reflects a growing trend in developer tools toward providing more "knobs and dials" for LLM interactions. By moving beyond simple chat interfaces and offering granular control over output length and image processing, Willison is catering to a technical audience that demands precision and efficiency in their AI-assisted development cycles.

While we strive for accuracy, bytevyte can make mistakes. Users are advised to verify all information independently. We accept no liability for errors or omissions.

AI-generated image.

✔Human Verified

Enhanced Vision and Custom Model Handling

Related Articles