Be careful: a lot of stuff got released since. I think GPT-4o has become the most economically viable strong model. Also I have seen people cheaply deploy custom models on runpod and modal plus (many others, see https://gpus.llm-utils.org/alternative-gpu-clouds/).
Maybe I should keep a constantly updating post with all of this :)
Thanks for your comparison, I was looking for something like that for Huggingface and ChatGPT but I couldn't find.
Again thank you so much !!
Be careful: a lot of stuff got released since. I think GPT-4o has become the most economically viable strong model. Also I have seen people cheaply deploy custom models on runpod and modal plus (many others, see https://gpus.llm-utils.org/alternative-gpu-clouds/).
Maybe I should keep a constantly updating post with all of this :)