Nvidia NIMS

For rapid prototpying, NVIDIA NIM inference microservices are a very good entry point to playing with LLMs. Two big advantages of the Nvidia models is that they are free, and they have a wide range of models available.

Setting up an account is simple. Head to the Nvidia Build website and sign up. If you use a personal account, you will get 1,000 free requests, and if you use a business account (or sign up for an enterprise trial, you will get 5,000 free requests).

You can now select from a broad range of models. Below is an example of the meta/llama-3.2-3b-instruct model. You can either play around with it in the Playground. Or you can use it as you would an OpenAI model.

Select Get API Key in the top right of the Python example, and you can get 24 hour access to the model. Once the 24 hours is up, you can always renew the key.