Summary & Key Takeaways ​

• llm 0.32rc2 switches the default model for new users to GPT-5.6 Luna.
• GPT-5.6 Luna is a more recent and capable model, albeit slightly more expensive than the previous default.
• A new llm openai endpoint command allows prompting arbitrary OpenAI-compatible endpoints.
• This new command supports prompts, chats, and model listings without prior configuration.
• It offers a convenient way to test local models like those in LM Studio.

Our Commentary ​

Simon Willison's llm tool continues to be an indispensable part of my AI workflow. The default model switch to GPT-5.6 Luna makes sense given the recent price drops, and the new llm openai endpoint command is a fantastic addition. I've been looking for a simple CLI to hit local OpenAI-compatible servers, and this fits the bill perfectly. It's these small, focused tools that often make the biggest difference in daily development.