llm 0.36
Simon Willison’s Weblog Simon Willison
llm 0.36 adds OpenAI’s gpt-6-sol and gpt-6-luna models. It also blocks chat-style prompts for single-turn models and hides reasoning traces in logs.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Simon Willison has shipped llm 0.36, and the headline is new model support from OpenAI: gpt-6-sol for GPT-6 Sol and gpt-6-luna for GPT-6 Luna. That’s the kind of update users notice right away, because it changes what the tool can talk to before they even touch a prompt.
The more interesting change is on the plugin side. Model plugins can now say supports_conversation = False when they only accept single-turn prompts. If llm gets assistant or tool history for one of those models, it raises llm.ConversationNotSupported instead of trying to force the issue. And llm chat now rejects those models before a session even starts.
That makes the failure mode cleaner, which is exactly the sort of boring-but-important polish that keeps a developer tool from becoming a small daily annoyance. The first plugin using it is llm-typesafe.
There’s also a tweak for logs: reasoning traces in Markdown output from llm logs are now wrapped in <details><summary> tags. So the long trace is still there, but tucked out of the way until someone wants to open it. The release also includes bug fixes from five new contributors.
My take — AI-written commentary, not fact-checked reporting
This is the right kind of release: less drama, more guardrails. The industry loves pretending every model can do everything, and then acts surprised when single-turn systems choke on chat history. A little honesty in the interface beats a lot of marketing glitter.
Read more about this at: Simon Willison’s Weblog
