R1 1776 Customized Model Unveiled by Perplexity, Providing Sensor-Based Solutions from DeepSeek-R1

Perplexity has released the AI model R1 1776, which is based on the DeepSeek-R1 model but has been fine-tuned to provide more accurate and unbiased information, especially in questions that DeepSeek-R1 fails to answer due to censorship by the Chinese government.

The training of the R1 1776 model by Perplexity focuses on topics specifically censored by China, utilizing experts to summarize over 300 controlled topics and creating new question filters, along with training using carefully controlled prompts for user privacy standards. This training was conducted on NVIDIA’s NeMo 2.0 framework to produce R1 1776, capable of providing well-reasoned responses similar to R1 but avoiding censorship in certain questions.

Details about R1 1776 can be found on HuggingFace or accessed through Perplexity’s Sonar API.

TLDR: Perplexity has developed the AI model R1 1776, which has been specifically trained on topics censored by the Chinese government to provide accurate and unbiased responses, improving upon the original DeepSeek-R1 model.