microFlare is a project to make small and medium sized LLMs more available to everyone. Even your older PC might be able to run these top 100 AI models, making this technology free for almost anyone to use, forever, and without having to pay a provider or take on all the risks of letting your private chats go into the cloud.
microFlare leverages asymmetric quantization, with precision chosen per layer to compress big models down to a size that can be ran on consumer PCs while maintaining similar quality to the original uncompressed LLM.
microFlare ver.1 can be ran on any PC with at least 16GB of RAM, and preferably with a 4GB or higher GPU if you want decent performance.
NanoFlare ver.1 can be ran on any PC with at least 10GB of system RAM, or completely on GPU if it has 8GB or more.
Contact: microFlare@proton.me