Qwen 2.5 72B Breaches Hard Prompts Top 10!!!

[@UCs6ktlulE5BEeb3vBBOu6DQ]

2.5 72B Q5 is my daily now. Coder 7B Q8 is also running for quick questions

[@Sean_neaS]

Personally I'd rather see examples than benchmarks. Pull it up on hugging chat or ollama and let me know what kind of CPU or GPU ii need. Maybe home much an api costs.

[@zyxwvutsrqponmlkh]

Always been a fan of the Qwen models

[@1littlecoder]

Highly underrated

[@nishanthvenkat5860]

How do you keep yourself updated with all these models on a daily basis. Any newsletter you subscribe to ? Or is there any other sources. Just curious

[@snapo1750]

i stopped using llama as soon as the qwen 2.5 32B model came out... with 2 GPU's i run it in q8_0 , the model is absolutely badass good....

[@1littlecoder]

Are using a multi GPU set-up?

[@alx8439]

I was able to run it on CPU only with 32 Gb of DDR5 RAM. IQ1 quant :) it was slow, yeah, but it works and still demonstrates a lot of intellect

[@winterclimber7520]

the Alibaba guys are unbelievable. i use agentscope literally every day, can't wait to try it with the qwen 2.5 models

[@xhridhar]

How’s Qwen 2.5 compared to Molmo ? Can you do a comparison please ?

[@TheRealUsername]

Hi, I think you should look up Liquid AI's LFM models, they are built with a novel type of neural net and they reportedly outperform Transformers.

[@dhruvmehta2377]

Have you tested the llama 3.2 90 billion?

[@ilianos]

(not finished watching this vid, but) it still can't answer all of my questions about a venn diagram

[@1littlecoder]

What kind of question did you ask with a Venn diagram?

[@hardtkk]

It sadly has a chinese political bias