Join our daily and weekly newsletters for the latest updates and exclusive content covering cutting-edge AI. Learn more
A few days ago, only the nerdiest nerds (I say that as such) had heard of it. Deep searchan equally evocative Chinese AI subsidiary High-level capital managementa quantitative analysis (or quant) company initially launched in 2015.
However, in recent days, it is undoubtedly the most discussed company in Silicon Valley. This is largely thanks to release of DeepSeek R1, a new large language model that performs “reasoning” similar to OpenAI’s current best available model o1 — taking several seconds or minutes to answer difficult questions and solve complex problems while reflecting on its own step-by-step analysis, or “chain of thought”.
Not only that, but DeepSeek R1 scored as high as or higher than OpenAI on various third-party tests (tests to measure AI performance in answering questions on various topics), and was reportedly trained at a fraction of the cost (around $5 million) with far fewer graphics processing units (GPUs) under a strict embargo imposed by the United States, OpenAI’s home territory.
But unlike o1, which is only available to paid ChatGPT subscribers of the Plus tier ($20 per month) and more expensive tiers (like Pro at $200 per month), DeepSeek R1 was released as a fully open source, which also explains why it quickly climbed the rankings of the AI code sharing community. Hugging Face’s Most Downloaded and Most Active Models.
Additionally, thanks to the fact that it is completely open source, users have already refined and trained many variations of the model for different task-specific purposes, such as making it small enough to run on a mobile device or combining it with other open source models. Even if you want to use it for development purposes, DeepSeek API costs are over 90% cheaper than the equivalent OpenAI o1 model.
The most impressive thing is that you don’t even need to be a software engineer to use it: DeepSeek has a free website And mobile app even for US users with an R1-powered chatbot interface very similar to OpenAI’s ChatGPT. Except, once again, DeepSeek has undermined or “mogged” OpenAI by connecting this powerful reasoning model to web search – something OpenAI has not yet done (web search is currently not available than on the least powerful GPT model family).
An open and closed irony
There’s a rather delicious, even disconcerting, irony to this given OpenAI’s founding goals of democratizing AI to the masses. As Jim Fan, head of research at NVIDIA, put it on X: “We live in a time where a non-US company is keeping OpenAI’s original mission alive: truly open, cutting-edge research that empowers all. This doesn’t make any sense. The funniest outcome is the most likely.
Or like User X @SuspendedRobot said itreferring to reports that DeepSeek appears to have been trained on Q&A results and other data generated by ChatGPT: “OpenAI stole the entire internet to get rich, DeepSeek stole it and gave it back to the masses for free. I think there’s a certain folk tale British on this subject”
But Fan isn’t the only one taking note of DeepSeek’s success. The open source availability of DeepSeek R1, its high performance and the fact that it seemingly “came out of nowhere” to challenge the former leader in generative AI, sent shockwaves throughout Silicon Valley and well into the world. beyond, based on my conversations and readings of various engineers, thinkers and leaders. If it’s not “everyone” who’s freaking out, as my hyperbolic title suggests, it’s certainly the most talked-about topic in tech and business circles.
A message posted on Blindthe anonymous gossip sharing app in Silicon Valley, has been circulating suggesting that Meta is in crisis because of DeepSeek’s success due to how quickly it surpassed Meta’s efforts to become the king of AI open source with its Llama models.
“It changes the whole game.”
User @tphuang wrote convincingly: “DeepSeek has made AI commonplace outside of the very high end. Lightbulb moment for me in the 1st photo. R1 is so much cheaper than the cost of labor in the United States that many jobs will be automated in the next five years. » later noting why DeepSeek’s R1 is more attractive to users than even OpenAI’s o1:
« 3 huge problems without o1:
1) too slow
2) too expensive
3) lack of end-user control/dependence on OpenAI
R1 solves them all. A company can purchase its own Nvidia GPUs and run these models. You don’t have to worry about additional costs or slow or unresponsive OpenAI servers. »
@tphaung also posted a compelling analogy as a question: “Will DeepSeek be to LLM what Android has become to the world of OS?”
Web entrepreneur Arnaud Bertrand also didn’t mince his words about the surprising implications of DeepSeek’s success: write on: “It cannot be overstated how profoundly this changes the entire game. And not just with regard to AI, it is also a massive indictment of the United States’ misguided attempt to stop the technological development of China, without which Deepseek might not have been possible (as the saying goes, necessity is the mother of invention).”
The question of censorship
However, others have raised cautions about DeepSeek’s rapid rise, arguing that as a startup operating out of China, it is necessarily subject to that country’s laws and content censorship requirements.
Indeed, my own use of DeepSeek on the iOS app here in the US revealed that this would be the case. does not answer questions about Tiananmen Squaresite of the 1989 pro-democracy student protests and uprising, and the subsequent violent crackdown by the Chinese military, leading to at least 200, if not thousands, deathswhich earned him the nickname “Tiananmen Square Massacre» in Western media.
Ben Hylak, former Apple human interface designer and co-founder of AI product analytics platform Dawn, posted on X comment asking about this topic caused DeepSeek R1 to enter a circuitous loop.
As a member of the press, I of course take freedom of expression extremely seriously and it is arguably one of the most fundamental and indisputable causes I champion.
Yet I would be remiss if I didn’t note that OpenAI’s models and products, including ChatGPT, also refuse to answer a whole host of questions about even innocuous content, particularly when it comes to human sexuality and erotic/adult topics, NSFW.
This is of course not an apples to apples comparison. And there will be some for whom resistance to relying on foreign technology will make them skeptical of DeepSeek’s ultimate value and usefulness. But we cannot deny its performance and low cost.
And at a time when 16.5% of all US products are imported by Chinait is difficult for me to warn against using DeepSeek R1 based on censorship concerns or security risks, especially when the model code can be downloaded for free, taken offline, used on the device in secure environments and tweaked at will. .
I certainly detect an existential crisis about the “fall of the West” and “rise of China”, which is however motivating some of the heated discussions around DeepSeek, and others have already linked it to this crisis.How US users joined the Xiaohongshu (aka “Little Red Book”) app when TikTok was briefly banned in that country, only to be amazed at the quality of life in China depicted in the videos shared there. The arrival of DeepSeek R1 occurs in this narrative context – one in which China appears (and by many indicators is clearly) ascendant while the United States appears (and by many indicators is also) in decline.
The first but not the last Chinese AI model to shake up the world
Nor will it be the last Chinese AI model to threaten the dominance of Silicon Valley giants – even if, like OpenAI, raise more money than ever for their ambitions to develop artificial general intelligence (AGI), programs that outperform humans at the most economically profitable work.
Just yesterday, another Chinese model from TikTok’s parent company Bytedance called Doubao-1.5-pro – was released with performance matching the GPT-4o model without OpenAI reasoning on third-party benchmarks, but again, at 1/50th the price.
Chinese models have become so good, so quickly, that even those outside the tech industry are taking notice: The economist the magazine has just published an article on the success of DeepSeek and other Chinese AI efforts, and political commentator Matt Bruenig posted on X that: “I have been using Gemini, ChatGPT and Claude extensively for NLRB document summarization for almost a year. Deepseek is better than all in this area. The chatbot version is free. The price of using its API is 99.5% lower than the price of OpenAI’s API. [shrug emoji]»
How is OpenAI responding?
It’s no wonder that Sam Altman, co-founder and CEO of OpenAI said today that the company brought its second family of reasoning models, o3, to ChatGPT, even for free users. OpenAI still seems to be paving its own path with more proprietary and advanced models, setting the industry standard.
But the question arises: with DeepSeek, ByteDance and other Chinese AI companies hot on its heels, how long can OpenAI stay ahead in creating and releasing new cutting-edge AI models? And if and when it falls, how deep and rapid will its decline be?
OpenAI, however, has another historical precedent. If DeepSeek and Chinese AI models do indeed become for LLMs like Google’s open source Android did for mobile – taking the lion’s share of the market for a while – just look at how the iPhone of Apple with its locked and proprietary all-in-one system The in-house approach has managed to carve out a place at the high end of the market and gradually grow downward from thereparticularly in the United States, to the point that it now holds nearly 60% of the national smartphone market.
Still, for anyone spending big money to use AI models from leading labs, DeepSeek shows that the same capabilities can be available at a much lower cost and with much greater control. And in a business setting, that can be enough to win.
