LLM Tool Reaches Version 0.31 with Official GPT-5.5 Model Support

Introduction: Another Major Iteration for Command-Line AI Tools

As the large language model ecosystem evolves rapidly, timely updates to developer tools are crucial. Recently, the open-source command-line tool LLM, maintained by renowned developer Simon Willison, officially released version 0.31, bringing native support for OpenAI's latest flagship model GPT-5.5 along with multiple enhancements to the model interaction experience. This update once again demonstrates LLM's vital role as a bridge connecting developers with cutting-edge AI capabilities.

LLM is a lightweight yet powerful command-line tool that allows users to interact with various large language models directly from the terminal, supporting models from multiple providers including OpenAI and Anthropic. It has earned a strong following among technical users thanks to its concise command syntax and highly extensible plugin architecture.

Core Updates: One-Command GPT-5.5 Access and Fine-Grained Control

Full GPT-5.5 Model Support

The most notable change in this update is undoubtedly the support for OpenAI's newly released GPT-5.5 model. Users simply need to type llm -m gpt-5.5 in the terminal to invoke the latest model directly, with no additional configuration required. This feature corresponds to GitHub Issue #1418 and marks another expansion of LLM's coverage of the OpenAI model family to the cutting edge.

As one of OpenAI's most powerful general-purpose models to date, GPT-5.5 delivers significant improvements in reasoning capability, multimodal understanding, and long-text processing. LLM's rapid adaptation allows developers to experience and test this new model's capabilities within their familiar workflows from day one.

Text Verbosity Control Option

Version 0.31 introduces a new text verbosity control option for GPT-5 and above models. Users can set the level of detail in model output via -o verbosity low, with selectable values of "low," "medium," and "high."

The practical significance of this feature should not be underestimated. In everyday use, developers frequently face a dilemma: sometimes they need only brief, concise answers, while other times they want thorough, in-depth analysis. The verbosity control option allows users to flexibly adjust output style based on specific scenarios, avoiding unnecessarily lengthy responses while still obtaining richer information when needed. This design reflects LLM's deep consideration of user experience details.

Image Detail Level Settings

For multimodal use cases, LLM 0.31 also introduces an image detail level (image_detail) setting option. When sending image attachments to OpenAI models, users can specify the image processing precision via -o image_detail low, supporting three standard values: "low," "high," and "auto." Notably, the GPT-5.4 and GPT-5.5 models additionally support an "original" option, allowing the model to process images at their original resolution.

This feature gives developers greater flexibility in balancing cost and quality. Selecting a lower image detail level can reduce token consumption and lower API call costs, while scenarios requiring precise image analysis can switch to high-precision or original mode for optimal results.

Async Model Registration Improvements

Additionally, this update fixes an important architectural issue (corresponding to Issue #1395): models custom-registered in the extra-openai-models.yaml configuration file are now automatically registered as async models as well. This means developers using custom OpenAI-compatible endpoints will enjoy a smoother experience during async calls and batch processing without additional manual configuration.

In-Depth Analysis: A Small Tool Reflecting Major Trends

On the surface, LLM 0.31 is merely a routine version update for a command-line tool, but it reflects several important trends in the current AI development ecosystem.

First, the pace of model iteration is accelerating. From GPT-4 to the GPT-5 series and now GPT-5.5, OpenAI's model release cadence has noticeably quickened. Whether downstream tools can keep up directly impacts developer experience and productivity. LLM's swift release of compatible versions following new model launches demonstrates the agile responsiveness of the open-source community.

Second, fine-grained control is becoming essential. As large model capabilities grow, simply entering a prompt and receiving a response no longer meets professional developers' needs. The emergence of options like verbosity control and image detail levels reflects users' dual focus on output quality and cost. In the future, we are likely to see more dimensions of control parameters introduced into various AI tools.

Third, command-line tools continue to gain prominence in AI workflows. Despite the proliferation of graphical interfaces and web applications, command-line tools retain irreplaceable advantages for technical users in scenarios such as automation scripting, CI/CD integration, and rapid prototype validation. LLM's continued active development is strong evidence of this trend.

Outlook: The Future Direction of Open-Source AI Toolchains

Looking ahead, open-source command-line tools like LLM still have vast room for growth. As multimodal model capabilities continue to strengthen, tool-level support for additional media types such as audio and video will become an inevitable direction. Meanwhile, as locally deployed open-source models mature, LLM's value in unified management of cloud-based and local models will become even more apparent.

For developers, the release of LLM version 0.31 is a noteworthy update milestone. Whether you want to experience GPT-5.5's capabilities at the earliest opportunity or seek finer model control within your existing workflows, this update delivers tangible value. Interested developers can quickly upgrade to the latest version via pip install --upgrade llm and unlock a more efficient AI interaction experience.