Article · 2026-07-30 · 4 moments

llm 0.32rc2

Release: llm 0.32rc2 Hot on the heels of RC1, this fixes a dependency issue and also adds two neat new features: The default model for users who have not set their own default is now GPT-5.6 Luna. It was previously GPT-4o mini. Luna is a much better and more recent model, albeit slightly more expensive - $0.20 per million input tokens and $1.20 per million output tokens, compared to $0.15/$0.60 for 4o mini. You can switch back to 4o mini using llm models default gpt-4o-mini, or switch ✦ AI generated

Read original ↗

GUEST Release author HOST Simon Willison

Mechanism◆

A new llm openai endpoint command lets users run prompts, chats, and model listings against arbitrary OpenAI-compatible endpoints without configuring a model first, and these calls are not logged

LLM 0.32rc2 introduces the llm openai endpoint command for interacting with any OpenAI-compatible endpoint without prior model configuration.

transcript

Release author: New llm openai endpoint command for running prompts, chats and model listings against arbitrary OpenAI-compatible endpoints without first configuring a model. These calls are not logged. #1565

gives example · 1provides context · 1

Fact◆

GPT-5.6 Luna is now the default model for llm, replacing GPT-4o mini

The default model for llm users without a custom default has changed from GPT-4o mini to GPT-5.6 Luna, offering better quality at a slightly higher price. Users can switch back to GPT-4o mini or to the cheaper GPT-5 nano.

transcript

Release author: The default model for users who have not set their own default is now GPT-5.6 Luna. It was previously GPT-4o mini. Luna is a much better and more recent model, albeit slightly more expensive - $0.20 per million input tokens and $1.20 per million output tokens, compared to $0.15/$0.60 for 4o mini. You can switch back to 4o mini using llm models default gpt-4o-mini, or switch to GPT-5 nano, an even cheaper default model ($0.05/$0.40), using llm models default gpt-5-nano. #1576

provides context · 2

Example

The llm openai endpoint command can be used as a uvx one-liner against an LM Studio local model without even installing LLM

The command works as a uvx one-liner, demonstrated with an LM Studio local model running google/gemma-4-31b, including tool support.

transcript

Release author: You don't even have to install LLM to use this. Here's a uvx one-liner for running a prompt - with tools - against an LM Studio local model: 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?' Output here.

Anecdote

The llm openai endpoint command was created out of frustration with the lack of an obvious CLI tool for testing prompts against arbitrary OpenAI Chat Completions imitation endpoints

The author built the feature because no simple CLI tool existed for trying prompts against arbitrary OpenAI-compatible endpoints.

transcript

Release author: The llm openai endpoint command is really cool. I got frustrated at the lack of an obvious CLI tool for trying out prompts against arbitrary OpenAI Chat Completions imitation endpoints, so I decided to add that to LLM itself.

Highlight slides

LLM 0.32rc2: New `llm openai endpoint` Command✦ from: A new llm openai endpoint command lets users run prompts, chats, and model listings against arbitrary OpenAI-compatible endpoints without configuring a model first, and these calls are not logged GPT-5.6 Luna Becomes Default Model✦ from: GPT-5.6 Luna is now the default model for llm, replacing GPT-4o mini Privacy: Calls Not Logged✦ from: A new llm openai endpoint command lets users run prompts, chats, and model listings against arbitrary OpenAI-compatible endpoints without configuring a model first, and these calls are not logged Input Token Pricing Across Models✦ from: GPT-5.6 Luna is now the default model for llm, replacing GPT-4o mini

Related episodes

Video · Syntax

LGTM, Ship It: The AI Code Review Problem

→

Video · Lenny's Podcast

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)

→

Video · 20VC

Zuckerberg Back on X Challenging Codex & Claude Code | SK Hynix’s $26BN IPO

→