lacaio

Well, I think you must agree that it makes sense to do it for science. Beyond stock markets and things like that, there is social and scientific interest in developing LLMs, or people wouldn’t do it. Ok, so let’s take a market kind of thing and say Microsoft only does Phi-4 and Phi-4-mini because of stocks and stockholders. Even if that translates into money, at the hollow core of this is social and scientific interest. So I think it makes much sense to have a Brazilian LLM, thank you.

Training LLMs is very costly, and open-weights aren’t open-source. For example, there are some LLMs in Brazil, but there is a notable case for a brazilian student on the University of Dusseldorf that banded together with two other students of non-brazilian origin to make a brazilian LLM. 4B model. They used Google to train the LLM, I think, because any training on low VRAM won’t work. It took many days and over $3000 dollars. The name is Tucano.

I know it looks cheap because there are many, but many country initiatives are eager on AI technology. It’s costly.

DeepSeek API isn’t free, and to use Qwen you’d have to sign up for Ollama Cloud or something like that, as Local deploying is prohibitive.

They’re trying to link DeepSeek to the old tale freeride companies that apparently have ties to the original company product and gets a “look the other way” attitude from it (e.g. Meta with their Whatsapp products). This situation is nothing like it.

Thank you for this.