LLM 0.37 adds GPT-6.1 models and a command for OpenAI-compatible endpoints

The release covers about two months of accumulated features and fixes. It adds gpt-6.1-sol and gpt-6.1-terra, plus an llm openai endpoint add command that makes it easier to configure additional models served from OpenAI-compatible endpoints.

The default model, used when a user has not set their own, changes to gpt-6-luna. Willison says it costs half as much as the previous default, gpt-5.6-luna. Embedding support now accepts that different binary content can have different content types, which models such as Gemini Embedding 2 need for text, audio, images and video.

Fixes came from 12 authors, almost all of them AI-assisted. Willison says reviewing them was less of a burden than he expected, and he treats many of the pull requests as bug reports with attached examples.

If your agents call models through this CLI or library, the default model is now gpt-6-luna, so check any scripts that rely on the old default. The new endpoint command lets you register OpenAI-compatible providers as extra models. Embedding calls can now pass binary content with different content types, which matters if you index audio or video.