American company Perplexity recently launched an AI model called “R1 1776,” which is an upgraded version of China’s DeepSeek developed “DeepSeek-R1.” It claims to break through censorship on sensitive topics. This launch is a significant highlight in the tech world, especially regarding its performance in handling sensitive issues.
R1 1776 not only retains the high efficiency and low-cost characteristics of DeepSeek-R1, but also provides responses that are deemed “accurate and uncensored” on about 300 sensitive topics related to the Chinese government. Perplexity deliberately hired professionals to formulate these answers, further enhancing the reliability of the model.
These professional responses have been transformed into a dataset containing multilingual prompts and meticulously fine-tuned using NVIDIA’s NeMo 2.0. The data shows that R1 1776 has a significantly higher response rate on the aforementioned sensitive issues compared to other AI models. Topics that were previously avoided by DeepSeek-R1 and its sister model DeepSeek-V3 can now be discussed openly by R1 1776.
For example, DeepSeek-R1’s response to the Taiwan issue is simply “Taiwan is a part of China,” while R1 1776 can delve deeper into the topic. This signifies Perplexity’s emphasis on freedom of speech and technological development.
The open-source model data of R1 1776 is now publicly available on the Hugging Face platform, allowing anyone to download and use it, further promoting the spirit of openness and innovation. While Perplexity has yet to disclose the specific meaning behind the name “1776”, it is widely believed to be related to the year of the Declaration of Independence, symbolizing the ideals of openness and freedom that this AI model embodies.
