Meta's recently released language model LLaMA, similar to ChatGPT, has been leaked and now this "shawwn/llama-dl" GitHub repository has instructions on how to download the 65B model weights (220 Gb total).
You can rent a GPU at Vast AI and run it at just $1.5/hour.
Other GPU rental sites - Lambda, FluidStack, CoreWeave - can charge up to 6 times more but that is still affordable.
