llm 0.37
- Source
- simonwillison.net
- Date
Context
Simon Willison's notes on llm 0.37, the new release of his command-line tool and Python library for working with large language models, include a report on a problem many open source maintainers complain about: a flood of pull requests written with AI help. is the project of his most affected by it, and he says it has not been a problem. The release includes fixes from 12 authors, almost all of them AI-assisted, and reviewing them was much less of a burden than he expected.
His main reason is that he has Anthropic's Opus 5.5 or OpenAI's GPT-6 Astra help him triage the submissions to find the most useful fixes. He also treats each one as a bug report with an attached example rather than code he must accept as written: sometimes he merges the contribution and sometimes he writes his own fix. That is one maintainer's experience with one project, not a survey of open source.
The release itself gathers about two months of features and fixes. It adds the gpt-6.1-sol and gpt-6.1-terra models and a command that makes it easier to configure other models served through OpenAI-compatible endpoints. Its support now accepts different content types for different kinds of binary input, which models such as Gemini Embedding 2, able to embed text, audio, images and video, need. For users who have not chosen their own default, the default model is now gpt-6-luna, which costs half as much as the previous default, gpt-5.6-luna.
- Simon Willison's llm 0.37 adds gpt-6.1-sol and gpt-6.1-terra, plus an `llm openai endpoint add` command for configuring other models served through OpenAI-compatible endpoints.
- For users who never set their own default, the default model changes to gpt-6-luna, which Willison says costs half as much as the previous default, gpt-5.6-luna.
- Embedding support now accepts different content types for binary input, which Willison says is needed for models such as Gemini Embedding 2 that embed text, audio, images and video.
- Willison says the flood of AI-assisted PRs has not been a problem for llm, mainly because he uses Opus 5.5 or GPT-6 Astra to help triage the most useful fixes. This release includes fixes from 12 authors, almost all AI-assisted.
- Willison treats AI-assisted PRs as bug reports with attached examples: sometimes he lands them, sometimes he implements his own fix instead.
- 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.
- multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
Checking sign-in…
Loading comments…