I have mixed feelings about it.

Like any new model, I gave it a varied set of questions from my ChatGPT history for comparison. It's suspicious that many paragraphs from its answers are extremely similar to ChatGPT's, only with a few words changed or rearranged. This makes me suspect it might be actually a frontier model stolen from some US big tech.

That being said, I definitely like talking to it more than ChatGPT. If it's stolen, they definitely did something to it to make it more likable. I find it better for actually talking to and drilling a subject (I learn science a lot with these models) because ChatGPT has been too expositionary recently.

I also think that it utilizes its web search ability better than ChatGPT. And I like that you can view its internal chain of thought that it produces before actually answering.

I bet it's a psyop. They didn't train it on $5m they either stole another opensource model and trained/finetuned on top of it or it was much more expensive amd trained with all GPUs bought in Africa and Thailand in the last 2 years 😂😂

Try asking the model about tienanmen square and see what happens. It’s an awesome Open Source LLM, but how a “restricted AI” for political reasons can be more advanced in the long term?