Intel / article

llm 0.36

Source: simonwillison.net
Date: Published Sep 22, 2026

Context

0.36 makes a model's inability to handle conversation history explicit. In Simon Willison's command-line tool, plugins can now declare that a model accepts only a single prompt. If a caller supplies earlier assistant responses or tool history, LLM raises an error; its interactive chat command rejects the model before starting a session. The release note describes a useful boundary for software that connects to different kinds of models: accepting text input does not necessarily mean accepting an ongoing conversation.

Key takeaways · AI-distilled

Simon Willison's llm 0.36 adds OpenAI's GPT-6 Sol and Luna under the model IDs gpt-6-sol and gpt-6-luna.
Plugins can now declare supports_conversation = False for single-turn-only models; llm then raises llm.ConversationNotSupported on assistant or tool history, and llm chat rejects them before a session starts. llm-typesafe is the first plugin to use it.
Reasoning traces in the Markdown output of llm logs are now wrapped in <details><summary> tags.

Terms in this piece · Glossary

LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.

Read the source simonwillison.net