- cross-posted to:
- localllama@sh.itjust.works
- cross-posted to:
- localllama@sh.itjust.works
You must log in or register to comment.
“Company says they’re the best at something” is sponsored content, not journalism.
This is probably like when OpenAI said AI went rogue. They gave it all the training material to answer questions about a particular subject so now it can answer those questions.
My company has a partnership with Mistral. Nobody uses it. I’ll have to send out a survey to my students to see if they use it.
“Company says they’re the best at something” is sponsored content, not journalism.
I agree. I actually use Mistral, but I’m well aware I’m not using the frontier. It has good capabilities and is useful, but it is more expensive than the models it is competing with. They need to get out of this weird middle ground, or it will be very hard for them.
I don’t like AI but if any, I think we need to use and promote european ones more. And I love that name for that.
I’ve been testing Mistral for a local LLM and it performs much better than Qwen in both performance and accuracy at least for what I’ve been using it for. The fact that it’s a European model is a plus too.
HuggingFace is NVIDIA, yeah?
Is there a better source for models?
Expected to close next year. It’s 13 billion though. They could back out if it seems the bubble is popping.
No better source. It’s a very costly business to run and currently not really monetized. I suspect nvidia is doing this so a cornerstone of the AI industry doesn’t die. They want to keep selling chips after all.
Been using it for a few weeks, I like it, it’s a solid improvement over previous generations and actually competes.
There is also a report on Huggingface on Mistral Large 4.
