llm 0.32rc2 has been released, arriving shortly after RC1. The update addresses a dependency issue and introduces two notable features for users of the CLI.
Default model change
For users who have not set a personal default model, the new default is GPT-5.6 Luna. The previous default was GPT-4o mini. Luna is a newer model and generally performs better, but it is somewhat more expensive: pricing listed in the release notes is $0.20 per million input tokens and $1.20 per million output tokens, compared with $0.15/$0.60 for GPT-4o mini.
Users can revert to the previous default with:
llm models default gpt-4o-mini
Or they can choose the cheaper GPT-5 nano as the default with:
llm models default gpt-5-nano
GPT-5 nano is listed at $0.05 per million input tokens and $0.40 per million output tokens.
(Source: release note #1576.)
New llm openai endpoint command
The release adds an llm openai endpoint command that lets you run prompts, chats, and model listings against arbitrary OpenAI-compatible endpoints without first configuring a model. Calls made with this command are not logged.
According to the developer note, the command was added because there was no obvious CLI tool for quickly testing prompts against OpenAI Chat Completions–style endpoints. The feature can be used even without installing llm itself: the notes include a uvx one-liner that queries a local LM Studio model (with tool support):
uvx --pre llm openai endpoint http://127.0.0.1:1234/v1
T llm_version -T llm_time --td
-m google/gemma-4-31b 'what is the current LLM version? And the time?'
The endpoint parameter can be any OpenAI-compatible URL; the example shows a local LM Studio endpoint.
(Source: release note #1565.)
Summary
Version 0.32rc2 provides a dependency fix, sets a newer default model (GPT-5.6 Luna) for unset users, and adds a convenient llm openai endpoint command to test OpenAI-compatible endpoints from the CLI. Pricing comparisons and commands to change defaults are included in the release notes.
