In November of 2022 when ChatGPT was released by OpenAI, I was tinkering with it on the first day. I remember a few years ago being on a fantasy football draft road trip in Colorado trying to show some of my friends how cool and easy it was to generate images with Midjourney. When OpenAI started letting individuals apply for GPT 4 API access, I was accepted within two weeks. There are plenty of people out there who are closer to the cutting edge than me but in terms of people I know in real life, I’ve been the one again and again freaking out over how mind blowing AI is while everyone I know shrugs skeptically.

That being . I was aware when LM Studio was released over the summer but generally uninterested. I use LLMs heavily at work and in my personal life and up until this point, the models I’ve had subscription access to have been far superior to anything from the open source community. However, recent fine tunes of LLAMA (NVIDIA AI Foundry’s in particular) are starting to come close to GPT 4 and Claude benchmarks. After some testing on HuggingFace I was convinced it was worth my time to start experimenting more with the open source LLM's.

I really can’t express how cool it is to be able to run something as powerful as Llama-3.1-Nemotron-70B-Instruct with perfect data privacy using nothing but the power of my gaming PC. In twenty years we’re going to look back on the last two years as the start of a technological revolution.