feedback
#3
by Disya - opened
I think Qwen / Qwen2.5-14B-Instruct-1M is seriously underrated. In my opinion, it follows instructions better than any other model of similar size and feels remarkably natural, especially in multilingual contexts—something that’s usually terrible in Mistral when it comes to RP.
The only downside is that it tends to get stuck on certain details more than other models do.
I don’t know what exactly Qwen changed, but Qwen2.5-14B-Instruct is significantly worse than Qwen2.5-14B-Instruct-1M.
My rating: 9.5/10. This is the model I almost always use.
Thank you so much for this well-tuned model
p.s. I’ve never encountered issues with Chinese characters (possibly because I always specify the language to write in).