Tool Reviews

LLM version 0.37 is now available.

LLM 0.37 is a consolidation release for the CLI and Python library. It adds gpt-6.1-sol and gpt-6.1-terra, changes the default model to the cheaper gpt-6-luna, and introduces llm openai endpoint add for simpler OpenAI-compatible setup. It also improves embeddings for mixed binary content and includes AI-assisted fixes from 12 contributors.

LLM version 0.37 is now available.

Key Takeaways

  • LLM 0.37 bundles around two months of features and fixes, including new `gpt-6.1-sol` and `gpt-6.1-terra` model support.
  • The default model changes to `gpt-6-luna` for users who have not set a custom default, and the release note describes it as half the price of the previous default.
  • A new `llm openai endpoint add` command simplifies configuring additional models that use OpenAI-compatible endpoints.
  • Embeddings support now handles mixed binary content types, improving workflows involving text, audio, images, and video.
  • The release includes fixes from 12 contributors, most of them AI-assisted, showing a pragmatic way to manage open-source contributions.

The LLM 0.37 release of the LLM command line tool and Python library bundles around two months of accumulated features and fixes. It adds gpt-6.1-sol and gpt-6.1-terra, introduces a new `llm openai endpoint add` command, changes the default model to gpt-6-luna for people who have not set their own default, and expands support for embeddings that handle different binary content types.

LLM version 0.37 is now available.

A cheaper default and simpler endpoint setup

The most immediate change for many users is the new default. The previous default was gpt-5.6-luna. In version 0.37, gpt-6-luna becomes the default when a user has not chosen another model. The release note says the new default is half the price of the previous default. Because no exact price figures are given in the source, the practical benefit is a lower-cost starting point rather than a fixed dollar amount.

The release also makes it easier to configure additional models that use OpenAI-compatible endpoints through the new `llm openai endpoint add` command. An OpenAI-compatible endpoint is a service interface that behaves like OpenAI's model endpoints. For readers who already connect the tool to such services, this turns a manual configuration task into a named command inside the project. The command is aimed at easier setup for additional models.

Embeddings now account for mixed binary content

LLM's embeddings support now accepts that different binary content may have different content-type values. The source describes this as necessary for models like Gemini Embedding 2, which can embed text, audio, images, and video. A content-type value is a label that tells a system what kind of data is being handled, so a tool that embeds a video file and a text file can distinguish the two.

This update matters for workflows that go beyond plain text prompts. If a user sends audio, images, or video into an embedding model through the library, the release note suggests that the project is preparing to handle the data more faithfully by preserving those content-type differences.

LLM version 0.37 is now available.

AI-assisted fixes from 12 contributors

The release includes fixes from 12 authors, and the note says almost all of them were AI-assisted. The maintainer says LLM is the project that has been most affected by the flood of AI-assisted pull requests that many open source maintainers complain about, but that this release did not feel as burdensome as expected. Opus 5.5 or GPT-6 Astra are used to help with triage and to spot the most useful fixes.

The contribution model described in the note is pragmatic. Each change is treated less like a large automated patch and more like a bug report with an attached example. Sometimes the maintainer lands the contribution, and sometimes the maintainer uses it as a signal and implements a separate fix. The credited contributors are 0d00ciallo0721, ac1982, Allen Flux, Bian Yuhan, Bishop Kammeraad, Chen Yufeiyang, Christopher Krah, Ethan Stoner, Jiachen Xie, Olamide Faniyan, Sharat Chandra Sai Boddu, and nosmile99.

What readers should take from the release

For people already using the LLM command line tool and Python library, version 0.37 is mainly a consolidation release. It does not announce a single large new product, but it brings several practical improvements together: access to gpt-6.1-sol and gpt-6.1-terra, a cheaper default path, easier configuration of OpenAI-compatible endpoints, and better handling of embeddings for mixed binary content.

LLM version 0.37 is now available.

The contributor list also shows how AI-assisted changes can be absorbed when triage is delegated to tools such as Opus 5.5 or GPT-6 Astra. That may be useful for open source maintainers who are worried about review overhead, though the note only describes the experience with this project. Readers who rely on a custom default model will not be forced into gpt-6-luna by this update. Readers who use embeddings with audio, images, or video, or who need to add OpenAI-compatible models, are the most likely to see a direct workflow benefit. The release notes do not provide exact pricing, so users should confirm costs before changing production defaults.

Release details

Meta description: LLM 0.37 adds gpt-6.1 models, a default switch to gpt-6-luna, endpoint configuration, richer embeddings, and fixes from 12 contributors in one release.

Tags: open-source, AI, embeddings, llms, annotated-release-notes

Featured image: abstract friendly illustration of connected model cards, file streams, and a terminal-style grid; no real people, logos, or text.

Comments (0)

  1. No comments yet. Be the first to share what worked for you.

Leave a comment

Comments are reviewed before they appear. Your email address is not published.