NVIDIA Nemotron-70B is good, not the best LLM

Though the model is good, it is a bit overhyped I would say given it beats Claude3.5 and GPT4o on just three benchmarks. There are afew other reasons I believe in the idea which I’ve shared here : https://youtu.be/a8LsDjAcy60?si=JHAj7VOS1YHp8FMV

3 Likes

No matter how much the hugging-face executives try to pump it up on Twitter, it’s nothing like the frontier models. To increase their own metrics for investors, they are basically pushing the “omg this is the next gpt4 at home” narrative every other week.

2 Likes

To be clear, I no longer evaluate things in relation to gptx. To me, closed source means nothing.

It’s nowhere if it’s not on huggingface.

1 Like

Why?

I would at least like to be conscious of the trade-off I’m making, even if I make a commitment to never employ closed-source models.

1 Like

Why?

I would at least like to be conscious of the trade-off I’m making, even if I make a commitment to never employ closed-source models.

1 Like

My place of employment expressly forbids access to openAI on a daily basis.

In addition, I continue to be asked to complete llm tasks.

Additionally, you cannot use the internet at all if you work in a secure workplace.

It doesn’t exist in my opinion if it isn’t open. I’m not alone, either.

1 Like

It makes sense. For a living, what do you do? Since it’s not the primary focus, I’m genuinely confused as to why someone is being requested to work on LLMs on the side without having access to OpenAI, especially through services like Azure OpenAI Service.