BestLLMfor

It takes an hour to download an open AI model and one second to find out it is too big for your graphics card. On BestLLMfor the machine comes first: you pick the graphics card or Mac chip, the memory, the system and what the model is for. The configurator then lists models that should run, with an estimated speed and the steps to install them.

Its suggestions come from a catalog of open models, 258 by the site's count, whose files anyone can download. Each entry shows the graphics memory a model needs at several levels of compression, which shrinks the files at some cost in quality. A separate calculator works out that memory from a model's size, and a comparison page sets two or three models next to each other.

The methodology page says the configurator's speeds come from a formula, while a benchmarks page lists figures measured on real machines. The memory check covers the model alone, so long conversations need spare room on top. Other apps can pull the same data from a JSON API, published under a CC BY 4.0 licence.