LLM 0.37: 12 Contributors and New GPT-6.1 Models

Version 0.37 of LLM, the command-line tool and Python library, is now available. It introduces new models, a command to add OpenAI-compatible endpoints, and an evolution of embeddings. The maintainer also details how they manage the influx of AI-assisted pull requests.

Twelve Corrections Integrated and AI-Assisted Sorting

Version 0.37 of LLM incorporates corrections proposed by 12 authors, most of whom the maintainer indicates were assisted by AI. They specify that the review workload was less than expected and that they treat these proposals as bug reports accompanied by examples, sometimes accepting them or applying their own corrections. To sort these contributions, they rely on Opus 5.5 or GPT-6 Astra to identify the most useful fixes. While many open-source project maintainers report an influx of AI-generated pull requests, the LLM lead believes their project is the most affected among theirs, without it being a problem. Twelve contributors are explicitly thanked for this version: 0d00ciallo0721, ac1982, Allen Flux, Bian Yuhan, Bishop Kammeraad, Chen Yufeiyang, Christopher Krah, Ethan Stoner, Jiachen Xie, Olamide Faniyan, Sharat Chandra Sai Boddu, and nosmile99.

Addition of gpt-6.1-sol, gpt-6.1-terra, and an Endpoint Command

Two new models, gpt-6.1-sol and gpt-6.1-terra, are now supported. A new command, llm openai endpoint add, simplifies the configuration of additional models via OpenAI-compatible endpoints.

New Default Model and More Flexible Embeddings

The default model is now gpt-6-luna, which is priced at half that of the previous default model, gpt-5.6-luna. Support for embeddings now allows for assigning different content types to various binary contents, a change deemed necessary for models like Gemini Embedding 2, capable of integrating text, audio, images, and video. All these changes are delivered in LLM 0.37, which encompasses about two months of new features and fixes.