LLM 0.32rc2: Enhancements and New Features Revealed

The latest update of LLM, version 0.32rc2, brings notable improvements in terms of features and bug fixes. Released on July 30, 2026, at 10:52 PM, it closely follows version RC1 and stands out with two major additions.

First, the default model for users who have not specified their own model is now GPT-5.6 Luna. This change replaces the previous default model, GPT-4o mini. GPT-5.6 Luna is a newer and more powerful model, although it comes at a slightly higher cost. The pricing is $0.20 per million input tokens and $1.20 per million output tokens, compared to $0.15 and $0.60 respectively for the 4o mini model. Users still have the option to revert to GPT-4o mini or choose GPT-5 nano, an even more economical model with costs of $0.05 per million input tokens and $0.40 per million output tokens.

The update also introduces a new command llm [openai](/dossier/openai) endpoint. This feature allows users to execute queries, discussions, and list models on endpoints compatible with OpenAI, without the need for prior configuration of a model. Calls made through this command are not recorded, providing increased flexibility for users.

In addition to these new features, the update fixes a dependency issue, thereby enhancing the overall stability and performance of the application. This command has been added to address the lack of an obvious CLI tool for testing queries against endpoints mimicking OpenAI Chat Completions. Users can even execute queries without installing LLM, thanks to a uvx command line that allows testing of local models from LM Studio.