Deepseek R1 Alternative

Nate

I'm not sure if anyone would be interested in an alternative DeepSeek R1 model, this one seems promising.

https://huggingface.co/perplexity-ai/r1-1776

Blog link: https://perplexity.ai/hub/blog/open-sourcing-r1-1776

R1 1776 is a DeepSeek-R1 reasoning model that has been post-trained by Perplexity AI to remove Chinese Communist Party censorship. The model provides unbiased, accurate, and factual information while maintaining high reasoning capabilities.

RoxyRoxyRoxy

Seems interesting but I’m not sure if it’s necessary. I’ve never had censorship issues with the R1 provider Kagi uses and running a quick test of some common likely-to-be-censored topics (including the one mentioned in the blog post) seems to give a pretty fair answer without any obvious censorship.

Vlad

The model provides unbiased, accurate, and factual information while maintaining high reasoning capabilities.

This is not possible by definition how LLMs work, I am afraid it is just marketing language.

yokoffing

It'd be cool to have some Qwen 2.5-72B derivatives, but they won't be available via API (as far as I know).

KagiFeedbackDuder

@Nate - I came here to say the exact same thing as you, as someone who is trying to leave Perplexity, and really enjoys the way this R1-1776 model's phrasing/communication style.

I'm open to the idea that with all my querying of the model, I likely haven't hit any "censorship" as some of the other folks in this thread have mentioned. Maybe I'm exaggerating the difference between the two in my mind. Thoughts?