LLM 0.36 adds GPT-6 models and a single-turn plugin flag

single source· 1 articles · confidence: medium · first seen 2026-09-22 18:48 UTC

What this means for you

Nothing to do unless you use llm or maintain a plugin for it. If the provider you support only takes single-turn prompts, set supports_conversation = False and the library will refuse conversation history up front rather than failing mid-session. 0.36 adds the ids gpt-6-sol and gpt-6-luna.

Simon Willison released llm 0.36, his command-line tool and Python library for calling models, on 22 September. It adds two OpenAI models — gpt-6-sol for GPT-6 Sol and gpt-6-luna for GPT-6 Luna — and lets a model plugin declare that it cannot hold a conversation. Such a model causes an error when handed assistant or tool history, and `llm chat` refuses it before opening a session; llm-typesafe is the first plugin to use this. Reasoning traces — the intermediate text a model produces before its final answer — in the Markdown output of `llm logs` are now folded into collapsible tags, alongside bug fixes from five new contributors.

Models in this story

Key facts

  • ·llm 0.36 was released on 22 September 2026 and adds the model ids gpt-6-sol (for GPT-6 Sol) and gpt-6-luna (for GPT-6 Luna). source
  • ·Model plugins can declare supports_conversation = False; the library then raises llm.ConversationNotSupported when handed assistant or tool history, and llm chat rejects the model before starting a session. source
  • ·llm-typesafe is the first plugin to use supports_conversation = False. source
  • ·Reasoning traces in the Markdown output of llm logs are now wrapped in <details><summary> tags. source
  • ·The release includes bug fixes from five new contributors. source

What the sources say

  • Simon Willison (press) — Release notes for the CLI tool: two OpenAI model ids, a single-turn plugin flag, folded traces.

Sources

The original reporting. Follow these — they did the work.

← the wire