DeepSeek, a Chinese Artificial Intelligence startup, has launched a new model that it claims is on par with OpenAI-o1.
In a recently released paper, the startup noted that its new flagship AI model, R1, showed a new level of reasoning. It tweeted, “DeepSeek-R1 is here. Performance on par with OpenAI-o1. Fully open-source model and technical report. MIT licensed: distill and commercialise freely.”
According to experts, this product is a huge leap in terms of scale and efficiency and may upend expectations of how much power and computing will be needed to manage the AI revolution, considering it was done on a cheaper budget than AI models like ChatGPT.
Read also: Investment in GenAI to rise by 60% in 3 years
“DeepSeek R1 is 100 percent open source and 96.4 percent cheaper than OpenAI o1 while delivering similar performance. OpenAI o1: $60.00 per 1M output tokens DeepSeek R1: $2.19 per 1M output tokens,” tweeted Shubham Saboo, a senior AI product manager in the US.
Arnaud Bertrand, founder of HouseTrip and Me & Qi, noted on X, “It’s essentially as if someone had released a mobile on par with the iPhone but was selling it for $30 instead of $1000.”
According to DeepSeek, R1 achieves performance comparable to OpenAI o1 across math, code, and reasoning tasks. Its research paper states that this is possible because of pure reinforcement learning. According to Jim Fan, senior research manager at Nvidia, this technique is why Google DeepMind’s AlphaZero is a master at games like Go and Chess from scratch.
DeepSeek, launched in 2023, explained in its paper that its goal is to explore the potential of AI to develop reasoning capabilities without any supervised data. It noted that an earlier version of R1, called R1-Zero, gave it an “aha moment” in which the AI “learned to allocate more thinking time to a problem to reevaluating its initial approach.”
“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive – truly open, frontier research that empowers all. It makes no sense. The most entertaining outcome is the most likely,” tweeted Jim Fan.
He noted that DeepSeek’s R1 is not only an open source but also spills all the training secrets of AI models. Bertrand of Me & Qi explained that the open source nature of R1 will enable plenty of AI applications that were unaffordable before.
He tweeted, “Say, for instance, that you want to build a service that helps people summarize books (random example). In AI parlance, the average book is roughly 120,000 tokens (since a “token” is about 3/4 of a word and the average book is roughly 90,000 words).
Read also: Artificial Intelligence (AI) –What impact will it have on education?
“At OpenAI’s prices, processing a single book would cost almost $2 since they charge $15 per 1 million tokens. DeepSeek’s API, however, would cost only $0.07, which means your service can process about 30 books for $2 vs just 1 book with OpenAI: suddenly your book summarizing service is economically viable.”
He noted that DeepSeek has changed the AI game. OpenAI began the AI race when it launched ChatGPT in November 2022. However, models from tech giants such as Google and Meta and those from across the world are catching up.
Join BusinessDay whatsapp Channel, to stay up to date
Open In Whatsapp
