Simon Willison's LLM CLI Ships 0.32rc2, Adds Support for New GPT-5.6 Model Family

Developer ToolsLLMs
Illustration generated by AI: Editorial image for Simon Willison's LLM CLI Ships 0.32rc2, Adds Support for New GPT-5.6 Model Family

The Core · TL;DR

  • LLM CLI reached version 0.32rc2 on July 30, 2026, fixing a dependency bug from 0.32rc1
  • 0.32rc1 added support for three new OpenAI models: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna
  • GPT-5.6 Luna is now the default model, priced at $0.20/M input and $1.20/M output tokens
  • New features include an 'llm openai endpoint' command for OpenAI-compatible APIs and a content-addressable hash ID schema for stored messages

Simon Willison's popular LLM command-line tool moved to version 0.32rc2 on July 30, 2026, patching a dependency bug that slipped into the previous release candidate just a day earlier. The fix keeps a fast-moving update cycle intact for a tool that has become a default way for developers to script and compare large language models from the terminal.

The bigger news arrived with 0.32rc1: support for three new OpenAI models, gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna. Of the trio, GPT-5.6 Luna has been set as LLM's new default model, replacing whatever previously held that slot and signaling where Willison sees the best balance of capability and cost for everyday use.

Pricing details clarify why Luna got the nod. It runs at $0.20 per million input tokens and $1.20 per million output tokens, positioning it above the ultra-cheap tier but still far below flagship pricing. For comparison, GPT-5 nano costs just $0.05 per million input tokens and $0.40 per million output, while GPT-4o mini sits at $0.15 and $0.60 respectively.

A New Storage Schema

Beyond model support, 0.32rc1 introduces a reworked internal schema for how LLM stores conversation history. Messages are now tracked using content-addressable hash IDs, a change that should make logs more reliable to deduplicate, audit, and reference programmatically.

The release also adds an llm openai endpoint command, letting users point the CLI at any OpenAI-compatible API rather than being limited to OpenAI's own hosted models. That's a meaningful extension for teams running local or self-hosted models behind an OpenAI-style interface, since it lets them reuse the same LLM workflows and plugins without switching tools.

Together, the two release candidates read as routine maintenance on the surface but carry real implications for how developers interact with rapidly multiplying model options. Keeping pace with OpenAI's expanding lineup, from nano to the 5.6 series, requires constant plumbing work, and Willison's quick RC1-to-RC2 turnaround shows the project treating that churn as a normal cost of staying current.

None of the sourced facts point to conflicting version numbers, dates, or pricing here: both release candidates and their details are corroborated across the same two changelog entries, dated identically to July 30, 2026.

Original reporting and research used to synthesize this article.

  1. 1llm 0.32rc2simonwillison.net
  2. 2llm 0.32rc1simonwillison.net
WK

WAKIB Editorial Team

This review was prepared and summarized by the WAKIB AI intelligence engine and vetted by our editorial board for accuracy and reliability.

Subscribe to Newsletter

Get a weekly summary of the most promising AI research and tools delivered to your inbox.

Telegram Channel

Join our active community on Telegram for real-time tracking of AI models and trends.

Join us on Telegram

More from Tools

View all in Tools