All the local LLM models I've ever tried are dumb as bricks like this. I can see why the paid ones are more popular I didn’t want my data to be harvested by companies such as OpenAI figured I’d put my 5080 to use and see how far LLMs have progressed since I last used it locally a few years ago. And well… most of my results were really bad like this one. Are Local LLMs always this terrible? They don’t seem to be able to search things on the internet either (Although I will be fair and say in other situations it’s cool & useful that it works completely offline) but it would be nice to toggle on/off internet access for cases like this, because clearly it didn’t know about the recompilation project or renderers I was talking about and comically hallucinated everything lol