Summary & Key Takeaways ​

• llm 0.37 introduces support for new models like gpt-6.1-sol and gpt-6.1-terra.
• A new llm openai endpoint add command simplifies configuring additional OpenAI-compatible models.
• The default model has been switched to gpt-6-luna, offering half the price of the previous default.
• Embeddings now support different binary content types, crucial for multimodal models.
• Simon Willison shares a positive experience with AI-assisted PRs, viewing them as valuable bug reports.

Our Commentary ​

Simon Willison's llm tool is a staple for many of us experimenting with LLMs locally. The cost reduction for the default model is a welcome change, and the improved embedding support is crucial as multimodal models become more common. His take on AI-assisted PRs is also refreshing; it's good to hear a positive perspective on how AI can genuinely assist in open-source maintenance, rather than just creating noise.