LLM Model Recommendations?

[joncursi]

Hello! I am keen to try this library and have read through all the docs. One thing that stuck out to me is that the model used throughout all of the docs is Opus 4.5 💰💰💰

I was wondering if you’ve done testing with less expensive models (eg GPT 4.1?) or if generative UI necessitates the use of coding-specific, power-house models like Opus.

Thanks!

[ctate]

Great callout. One of the best parts is that this works really well with smaller models

I've turned it into an env var AI_GATEWAY_MODEL for both the app and the example. It defaults to anthropic/claude-haiku-4.5

You can see the full list of supported models here (including openai/gpt-4.1) here:

https://vercel.com/ai-gateway/models

[gaogaomie]

I've read the documentation and noticed that json-render currently seems to only support integration with the Vercel AI SDK for calling AI models. In some work scenarios, we can only use locally deployed large language models. Are there any plans to add support for calling locally hosted models in the future?