The default model changes to GPT-6 Luna for users without their own default, which can alter behavior and cost; llm models default gpt-5.6-luna reverts it. Embedding plugins can now accept typed attachments such as images.
Key takeaways · AI-distilled
Before 0.37, LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → chat logged every response even after logging was turned off with llm logs off; it now respects that setting.
Response.reply() now passes before_call and after_call to the tool calls it runs automatically. Previously a tool that before_call rejected would still run.
Collection.embed_multi() now raises ValueError when a model returns the wrong number of embeddings for a batch, instead of silently losing rows or storing vectors against the wrong IDs.
The release adds Python 3.15, drops Python 3.10, and removes shut-down OpenAI models including gpt-4, gpt-4-turbo, o1, o3-mini and o4-mini, along with aliases such as 3.5, chatgpt, 4 and 4t.
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.
embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
