Deepseek releases updated R1 model with improved AI performance and fewer hallucinations

[Redaktion]

DeepSeek has announced DeepSeek-R1-0528, the latest version of its R1 AI model, featuring improved reasoning, reduced hallucinations, and enhanced performance across multiple AI benchmarks, positioning it closer to top proprietary models.

https://www.notebookcheck.net/Deepseek-releases-updated-R1-model-with-improved-AI-performance-and-fewer-hallucinations.1026971.0.html

[majestictodos]

The funniest thing about Deepseek is when you ask it questions about the country of Taiwan, and you get some of the most unhinged responses lol

[Dave10]

I guess Grok is the best, although it's not shown in the chart. Deepseek does not respond to politically sensitive questions especially relating to China, while Open-AI is "Closed" in countries deemed adversaries of the USA. And this makes Grok the real "Open" AI in the market.

[Spectroscopy]

Great model.
If you could just overcome the fact that its "the most censored DeepSeek model yet for criticism of the Chinese government."
Source "DeepSeek's updated R1 AI model is more censored, test finds", tc