What is DeepSeek and how is it outdoing ChatGPT?

Its cost-efficient model and speedier processing makes it a compelling alternative to industry leaders

Chinese company DeepSeek is causing quite a stir in the artificial intelligence world.

With its free-to-use and cost-efficient model, the company's large language models are emerging as a serious challenger to ChatGPT and other generative AI platforms.

Here, we break down what you need to know about the new AI kid on the tech block.

How good and fast is DeepSeek?

DeepSeek's latest platform, introduced last week, is on a par with the fields top-tier models, including OpenAI's ChatGPT o1 and Anthropic's Claude 3.5 Sonnet. Proponents say it is better.

A key aspect in which DeepSeek has others beat is its speed. According to tests, it is nearly twice as fast as the top models. DeepSeek can process 60 tokens a second, compared to the 20 and 18 from GPT-4 and Meta Platforms' Llama, respectively.

Tokens in AI are the basic unit to calculate the length of text and can include punctuation marks and spaces, and vary from language to language. The bigger the token count, the more comprehensive the generated results are.

Is DeepSeek really free?

As it stands, DeepSeek is available anywhere in the world on the web and through app stores.

The free versions require a user to register by email, Google or a Chinese phone number, starting with the prefix +86, supported wherever you are.

Advanced users like developers who want to, for example, build chatbots, must pay for tokens.

But DeepSeek is really cost-efficient: you have to pay just $0.14 and $0.55 per million tokens for its V3 and R1 models, respectively, compared to ChatGPT's $15.

Top generative AI models have free services, but that won't give you access to their most premium features, such as higher creative output, mathematical equations and logical reasoning – things that DeepSeek offers.

Altogether, these higher level features attract users who want greater capabilities from their large language models without having to pay US prices for them.

DeepSeek’s entry “has already led to a shift in user sentiment as its more affordable solutions are quickly gaining popularity, drawing users away from established players … suggesting a potential shift in the competitive landscape”, Hassan Fawaz, chairman and founder of UK-based broker GivTrade, told The National.

It is unclear how many DeepSeek users there are. ChatGPT has more than 300 million weekly users, OpenAI chief executive Sam Altman said in December.

However, DeepSeek has exceeded ChatGPT as the top downloaded productivity app on the Apple App Store. Interest is so high that DeepSeek cannot be downloaded because of overwhelming demand and number of users. Its site was hit by a cyber attack on Monday blamed on the product's high profile.

How is DeepSeek able to offer its platform for free?

The simplest answer is by using a cost-efficient model. In other words, spending significantly less than the biggest names in generative AI.

Compared to OpenAI and Meta, who have been spending anywhere from tens of millions to billions training their large language models, DeepSeek was able to do the only with a reported $6 million, and in a significantly smaller time frame.

DeepSeek has overtaken ChatGPT on Apple's App Store. Reuters

Although not officially claimed by DeepSeek, logic dictates that the company is able to offer its large language models for free as it does not need to pass on the costs of training and development to consumers. It is possible that the revenue they generate from paid tokens could be enough to fund them.

“By developing cutting-edge AI models with less advanced and more cost-efficient hardware, DeepSeek challenges the heavy investments US tech companies are pouring into high-cost AI infrastructure,” said Charu Chanana, chief investment strategist at Danish lender Saxo Bank.

It is unclear what DeepSeek's plans are, and it will not be a surprise if the company starts offering paid tiers.

Is DeepSeek safe?

At least one cyber security firm has voiced concerns about vulnerabilities on DeepSeek.

Kela, an Israel-based company, claims that DeepSeek can be “easily exploited”, and that it was able to jailbreak its model “across a wide range of scenarios, enabling it to generate malicious outputs, such as ransomware development, fabrication of sensitive content, and detailed instructions for creating toxins and explosive devices”.

DeepSeek has not responded to Kela's report.

What are the big names saying about DeepSeek?

DeepSeek's rise – it has even unseated ChatGPT on Apple's App Store over the weekend – caught everyone by surprise, prompting acknowledgement from key personalities.

US President Donald Trump said that DeepSeek is a “wake-up call” for US companies to do better. Mr Altman described it as “impressive” promising OpenAI will raise their game, while Meta, meanwhile, reportedly set up “war rooms” to counter DeepSeek's rise.

Meanwhile, AI chip leader Nvidia also gave DeepSeek compliments, calling R1 an “excellent AI advancement”. Its chief executive Jensen Huang, however, lost about $20 billion from his wealth, thanks to a tech sell-off triggered by DeepSeek's rise.

How much has DeepSeek wiped off markets?

With a sell-off in technology stocks on Monday, stocks and the wealth of tech bosses were hit hard. The Nasdaq Composite on Wall Street alone lost about $1 trillion on Monday, with Nvidia alone making up about $600 billion of that.

The wealth of tech moguls mostly tied to AI – Mr Huang, Oracle's Larry Ellison, Dell's Michael Dell and former Binance chief Changpeng Zhao – were the biggest losers, losing nearly $68 billion on Monday, according to the Bloomberg Billionaires Index.

“Nvidia's slump highlights the complexities of a constantly evolving tech market, where disruptive advancements can reshape the landscape in a few days,” said Antonio Di Giacomo, a senior market analyst at XS.com.

Even energy stocks felt the heat, going lower on fears that DeepSeek may cause a shift towards using less energy in running AI platforms.

Will DeepSeek reshape investment sentiment?

DeepSeek's cost-effective model has cast doubts over the splashy and expensive methods US companies employ in developing AI models.

The company is poised to challenge Silicon Valley’s leadership, disrupt the global tech landscape and reshape the direction of the AI arms race, said Nigel Green, chief executive of global financial advisory firm deVere Group.

“US Big Tech, long considered untouchable, is facing credible and intensifying competition from China. For investors, this is both a warning and an opportunity; it’s time to rethink traditional tech allocations and seek out new areas of growth,” he said.