The top line
Here’s everything to know about DeepSec, the Chinese AI company that topped the app charts and shook up global tech stocks on Monday when it outperformed its top US rivals. Classified.
This photo illustration shows the DeepSec app on a mobile phone in Beijing on January 27, 2025.
Key facts
The DeepSec startup is less than two years old — it was founded in 2023 by 40-year-old Chinese entrepreneur Liang Wenfeng — and released its open-source models for download in the United States in early January, where It has grown since then. OpenAI’s Chat topped the iPhone download charts, overtaking the app for GPT.
DeepSeek’s latest product, an advanced inference model called R1, has been compared to the best products from OpenAI and Meta while to appear To be more efficient, the models are built with less cost to train and develop and without relying on the most powerful AI accelerator possible. Hard to buy Because of US export controls in China.
Both the company’s R1 and V3 models are in the top 10. Chatbot Arenaa performance platform hosted by the University of California, Berkeley, and the company says it’s nearly out-scoring rival models in math tasks, general knowledge and question-and-answer performance. Benchmarks.
DeepSec said One of its most advanced models cost $5.6 million to train, far less than an AI chief executive’s $100 million to $1 billion. Estimated Costs to build a model last year — though Bernstein analyst Stacey Rasgan later called the DeepSec figures highly misleading.
Despite lingering questions about the true cost and process of building DeepSec’s products, they still sent the stock market into a panic: Microsoft (down 3.7% as of 11:30 a.m. ET), Tesla (1.3%), Nvidia (15%) and Broadcom (16%) all slipped on Monday after a successful selloff from the DeepSec, and the tech-heavy Nasdaq was down. 3.5% on track for its third-worst day in the past two years.
Get Forbes Breaking News text alerts: We’re launching text message alerts so you’ll always know the biggest stories shaping the day’s headlines. Text “Alerts”. (201) 335-0739 Or sign up. Here.
Important quote
“Deepseek R1 is AI’s Sputnik moment,” billionaire investor Marc Andreessen said Sunday.
Big number
1.6 million is how many times the DeepSec mobile app was downloaded as of Saturday, Bloomberg Reportedly, the No. 1 app in iPhone stores in Australia, Canada, China, Singapore, the US and the UK
How do I use DeepSec?
DeepSeek is free to use on the web, app and API but requires users to create an account.
How does DeepSec compare to OpenAI and ChatGPT?
DeepSeek-R1 is most similar to OpenAI’s o1 model, which costs users $200 per month. Both are large language models with advanced reasoning capabilities, different from short-form question-and-answer chatbots like OpenAI’s ChatGTP. R1 and o1 specialize in splitting requests into a chain of logical “thoughts” and testing each one individually. The R1 has achieved similar performance to the o1 in several benchmarks. Allegedly His performance on the MATH-500 test exceeds that. Chatbot Arena is currently tied with R1 for the third-place AI model, with o1 in fourth.
Is Deepseek-R1 open source?
Yes DeepSeek-R1 is available for anyone to access, use, study, modify and share, and is not restricted by proprietary licenses.
Who owns DeepSec?
DeepSeek operates independently but is funded solely by High-Flyer, an $8 billion hedge fund also founded by Wenfeng. The company released its first product in November 2023, a model designed for coding tasks, and subsequent releases, notable for their low cost, have rivaled other Chinese tech giants. Forced to lower prices of their AI models to stay afloat. In one The interview Last year, Wenfeng said the company does not aim to make excessive profits and that its products are priced slightly above their cost.
Is DeepSec really that cheap?
Not everyone is buying the claim that DeepSeek built the R1 on a shoestring budget and without the aid of American-made AI chips. SCALE AICEO Alexander Wang DeepSeek told CNBC on Thursday (without evidence) that DeepSeek built its product using about 50,000 Nvidia H100 chips, which cannot be listed because it would violate US export controls that restrict sales of such chips to Chinese companies. Banning. Billionaire Elon Musk supported the theory, saying it “Obviously“True in a post on X. Rasgan echoed similar sentiments in a note on Monday, writing that DeepSec did not quantify the resources used to develop the R1 model in its reports and that that “the models look fantastic… He also said that the $5 million cost estimate may accurately represent what DeepSec has to rent the dedicated infrastructure to train its models. What did you pay for? But it excludes prior research, experiments, algorithms, data and costs associated with building its products.
What to watch for?
How much success will Nvidia, maker of the most sought-after artificial intelligence chips, get? The company is tracking for an 11%, or $400 billion, loss, which would be the biggest single-day value loss ever for any company. That record is already held by Nvidia, which fell nearly 10 percent in September to lose $280 billion in market value.
Key background
China and the US are major players in the artificial intelligence arms race, which appears to be led mainly by US firms OpenAI (backed by Microsoft), Meta and Alphabet. Last week, President Donald Trump backed $500 billion of OpenAI. Stargate Infrastructure plans to outdo its peers and specifically announce its support. said About the importance of US dominance over China in the AI ​​space. Artificial intelligence is largely driven by high-tech and high-dollar semiconductor chips that provide the processing power needed to perform complex calculations and efficiently handle large amounts of data. And while not all major semiconductor chipmakers are American, many—including Nvidia, Intel and Broadcom—are designed in the United States. In 2022, the US started limiting Semiconductor exports to China in an attempt to curb the country’s ability to advance AI for military applications or other national security threats.
Further reading
