Chinese AI lab DeepSeek has released an open version of DeepSeek-R1, its so-called reasoning model, which it says performs just in addition to OpenAI's o1 on certain AI benchmarks.
R1 is out there on the AI development platform Hugging Face under an MIT license and may due to this fact be used commercially without restrictions. According to DeepSeek, R1 outperforms o1 on the AIME, MATH-500 and SWE-bench Verified benchmarks. AIME uses other models to judge a model's performance, while MATH-500 is a group of word problems. SWE-bench Verified, however, focuses on programming tasks.
Because it’s an argumentation model, R1 effectively checks the facts itself, avoiding a number of the pitfalls that typically trip up models. Reasoning models take a little bit longer – normally seconds to minutes longer – to reach at solutions in comparison with a typical non-reasoning model. The advantage is that they have an inclination to be more reliable in areas reminiscent of physics, science and arithmetic.
R1 comprises 671 billion parameters, DeepSeek revealed in an announcement Technical report. Parameters roughly correspond to a model's problem-solving capabilities, and models with more parameters generally perform higher than those with fewer parameters.
671 billion parameters is huge, but DeepSeek has also released “distilled” versions of R1, ranging in size from 1.5 billion parameters to 70 billion parameters. The little ones can run on a laptop. The full R1 requires more powerful hardware, but is out there via DeepSeek's API at prices 90-95% cheaper than OpenAI's o1.
R1 has an obstacle. Since it’s a Chinese model, it’s subject to Benchmarking from China's web regulator to make sure their responses “embody fundamental socialist values.” For example, R1 doesn’t answer questions on Tiananmen Square or Taiwan's autonomy.
Many Chinese AI systems, including other reasoning models, refuse to answer topics that might arouse the ire of regulators within the country, reminiscent of speculation about them Xi Jinping Regime.
R1 arrives days after the outgoing Biden administration proposed it harder Export rules and restrictions on AI technologies for Chinese corporations. Companies in China were already blocked from purchasing advanced AI chips, but when the brand new rules take effect as written, corporations will face stricter caps on each the semiconductor technology and the models needed to launch advanced AI systems are.
In a policy document last week, OpenAI called on the U.S. government to support the event of U.S. AI in order that Chinese models don’t match or exceed it in performance. In one interview With The Information, Chris Lehane, vp of policy at OpenAI, highlighted High Flyer Capital Management, DeepSeek's parent company, as a corporation of particular concern.
So far, a minimum of three Chinese labs – DeepSeek, Alibaba and Howwhich is owned by Chinese unicorn Moonshot AI, has produced models that they are saying compete with o1. (It's value noting that DeepSeek was first – it announced a preview of R1 in late November.) In a post On
“The impressive power of DeepSeek’s distilled models (…) implies that very powerful thinkers will proceed to proliferate and have the opportunity to run on local hardware,” Ball wrote, “removed from the eyes of a control regime from above.”
