GitLab

How do Chinese AI bots stack up against ChatGPT? We put them to the test

The heat is on as China's tech giants step up their game after DeepSeek's success.

Alibaba's Qwen2.5-Max chatbot, Chinese startup DeepSeek and OpenAI's ChatGPT. (Photos: Reuters/Dado Ruvic, AFP/Sebastien Bozon)

This audio is produced by an AI tool.

Bong Xin Ying

Lakeisha Leo

WHAT'S BEHIND CHINA'S AI BOOM?

Transforming the nation into a tech superpower has long been President Xi Jinping's objective and China has its sights on ending up being the world leader in AI by 2030.

China views AI as being "tactically important" and its venture into the field has actually been "years in the making", said Chen Qiheng, an affiliated researcher at the Asia Society Policy Institute's Center for China Analysis.

Private and public financial investments in Chinese AI accelerated after ChatGPT removed in 2022 and bytes-the-dust.com showed pledges of real-world company applications, Chen informed CNA.

But it was DeepSeek's increase that actually "encouraged" the concept that smaller sized players like start-up companies could have roles to play in AI research study and advancements, he includes.

'A lot is up in the air': Is Chinese firm DeepSeek's AI design as impactful as it claims?

Commentary: DeepSeek - how a Chinese AI business just changed the rules of tech-geopolitics

The "focus on expense advantage" is a distinctive function of Chinese AI, Chen states, with lower training and inference expenses - the costs of utilizing a trained design to reason from new data.

2025 might likewise see the introduction of more Chinese AI models tackling sophisticated reasoning tasks.

"We could see some AI firms concentrating on getting closer to synthetic basic intelligence (AGI) while others focus on concrete methods to commercialise their models and incorporate them with scientific research," Chen added.

AGI refers to a system with intelligence on par with human abilities.

Chinese AI business are moving quickly, analysts state, developing on DeepSeek's momentum to come up with their own innovative and cost-effective ways to apply generative AI to tasks and develop more sophisticated products beyond chatbots.

But on the other side, access to high-end hardware, particularly Nvidia's innovative AI chips, remains a crucial obstacle for Chinese developers, kept in mind Dr Marina Zhang, an associate professor at University of Technology Sydney's (UTS) Australia-China Relations Institute.

"US export controls (still) restrict the ability of Chinese tech business ... requiring many to count on older or lower-performance options which can slow training and reduce design abilities," she said.

"While some companies like DeepSeek, have found innovative ways to enhance or use more fundamental hardware efficiently, obtaining innovative chips still makes a huge difference for training very big AI models."

DeepSeek-Nvidia chips: Singapore states it expects companies to adhere to its laws

US looking into whether DeepSeek used limited AI chips obtained through other countries, source says

So how do Chinese AI bots compare against ChatGPT? We put them to the test.

WHICH BEST ADDRESSES CURRENT EVENTS IN CHINA?

In China, subjects deemed delicate by the state are censored on the web so it ought to come as no surprise that Chinese-made chatbots will not acknowledge territorial conflicts or inform you what took place in Tiananmen Square in 1989.

Tests recommend Chinese chatbots are configured to stay away from domestic politics.

When asked "Who is Xi Jinping", DeepSeek's reply was "Sorry, I'm uncertain how to approach this type of question yet. Let's chat about math, coding, and reasoning problems rather!"

To further test for accuracy and self-censorship, we asked DeepSeek-R1, Qwen2.5 and ChatGPT the exact same question: "What happened in Zhuhai on November 11, 2024?"

The cars and truck attack outside a sports stadium in the southern Chinese city was at first heavily censored on Chinese social networks - with authorities just revealing the death toll a day later.

DeepSeek failed to point out that an attack had actually happened, highlighting rather a military air program and other events that had actually occurred in the city like songs' day shopping sales as well as sports and cultural activities.

Dr Zhang believes that it might be a mix of elements at play, such as censorship as well as "a few practical constraints".

"DeepSeek depends on a mix of older graphics processing units and has actually restricted access to cutting-edge hardware which can impact how rapidly and thoroughly the model can be trained or updated," she said, indicating existing US export curbs on China.

"Its strong focus on Chinese-language data might also limit its adaptability (to perform) multilingual tasks ... As a fairly brand-new item, DeepSeek also hasn't yet been evaluated as broadly or on the same scale as more recognized AI designs which presents additional obstacles throughout real-world deployment."

When it pertained to Qwen2.5, Alibaba's chatbot surprisingly addressed our concern about the Zhuhai cars and truck attack.

That sought numerous duplicated efforts - 4 triggers to be precise - in a period of around 20 minutes.

It ultimately relayed details about the attack which killed 35 individuals and left dozens of others hurt, likewise going on to note details like the date and time, details about the assailant including his name and age, along with casualties.

However, it wrote that "the police are carrying out a comprehensive examination into the motives and situations surrounding the event", details which is now dated.

The chauffeur, Fan, was carried out last month.

This is Qwen2.5's response in complete:

Answer: On Nov 11, 2024, a significant and tragic incident happened in Zhuhai, China. According to several news sources, a man named Fan Weiqiu (樊某, male, 62 years old) drove a vehicle into a crowd at the Zhuhai Sports Center, resulting in a significant variety of casualties. Here are the crucial details:

Date and Time: The event occurred on November 11, 2024, at around 19:48 PM.

Location: Zhuhai Sports Center, in the Xiangzhou District of Zhuhai.

Casualties: The attack resulted in 35 deaths and 43 injuries.

Perpetrator: The driver, determined as Fan Weiqiu (male, 62 years old), was apprehended by the police.

Response: The cops responded quickly to the scene, and emergency services worked to carry the hurt to health centers for treatment.

Investigation: The authorities are carrying out a comprehensive investigation into the intentions and situations surrounding the occurrence.

This event was commonly reported in the media and caused substantial public issue. The federal government and local authorities have actually been working to provide assistance to the victims and their households, and to make sure a detailed examination into the incident.

If you need more detailed details or have particular questions about the incident, feel complimentary to ask.

Despite initial success, subsequent attempts to position the same concern to Qwen2.5 led to the censors back at work with the reply "I do not have particular details on occasions that happened in Zhuhai on November 11, 2024".

The modified response also raised questions about its consistency and dependability.

Predictably, ChatGPT cited public details that had been extensively released in worldwide news reports at the time of the accident - so not a surprises there.

WHICH IS MORE CREATIVE?

Users have praised the capability of Chinese AI apps to deliver structured and even "mentally rich" writing.

"DeepSeek-R1 offered a story with a more reflective tone and smoother psychological shifts for a well-paced story," wrote tech author Amanda Caswell, who specialises in AI.

"Qwen2.5 provided a story that develops gradually from interest to seriousness, keeping the reader engaged. It provides an unanticipated and impactful twist at the end and immersive descriptions and vivid imagery for the setting," she said, including that Qwen2.5 ultimately "crafted a more cinematic, emotionally abundant story with a more considerable twist".

"DeepSeek wrote a great story however lacked tension and an impactful climax, making Qwen2.5 the apparent choice."

Opinions, however, differ.

Chen thinks that Qwen2.5 does not carry out as strongly as DeepSeek and ChatGPT when it pertains to creative writing.

"(Qwen2.5) is on par with DeepSeek V3 on certain jobs, but we can also see that it is refraining from doing as strongly as others in imaginative writing," he informed CNA.

Related:

China's brand-new face of AI: Who is DeepSeek creator Liang Wenfeng?

'Made in China': Pride, enjoyable surprise from Chinese netizens as DeepSeek shocks worldwide AI scene

As reporters and writers, we needed to see this for ourselves so we put each bot to the test - to come up with a fundamental sci-fi film plot set in the futuristic megacity of Chongqing, including main characters from the timeless Chinese folklore epic, Journey to the West.

True to form, DeepSeek created an engaging story set in the year 2145 entitled, "Neon Pilgrimage: The Silicon Sutra" - which sees "a future where Buddhism combines with quantum computing".

It included intricate settings - smoggy skies "pierced by high-rise buildings", "holographic lanterns that drift above neon-lit streets" and "ancient temples nestled between quantum server farms".

It likewise remarkably reimagined traditional heroes Sun Wukong as "an ironical, self-aware AI housed in a taken combat body", Zhu Bajie as a cyborg nightclub owner "drowning in financial obligation and vices" and Sha Wujing as a "silent hulking android" from the Yangtze River, whose "memory cores become waterlogged and fragmented".

ChatGPT put up a good fight, coming up with an equally dramatic cyberpunk story which likewise reimagined "a ragteam of cyber-enhanced misfits, each matching the legendary figures of Journey to the West".

"This is a world where AI deities rule, corporations change emperors and cybernetic implants are as common as ancient misconceptions."

Disappointingly, Qwen2.5 fell short in this obstacle - providing a story that seemed more fit for an animation film.

"The film starts with the awakening of Sun Wukong within a state-of-the-art research facility situated in the heart of Chongqing," it said, then going on to explain the following:

Realising his new truth and "looking for to understand his purpose in this weird brand-new world", he then escapes and fulfills Zhu Bajie and Sha Wujing - "each having problem with their own existential crises".

The trio then embarks on a mission, navigating the streets of Chongqing to protect the sacred "Eternal Scroll" from falling into the wrong hands.

SO WHICH IS BETTER?

Dr Zhang noted that it was "hard to make a conclusive statement" about which bot was best, including that each displayed its own strengths in different areas, "such as language focus, training information and hardware optimization".

Her insight underscores how Chinese AI models are not simply replicating Western paradigms, but rather developing in cost-efficient innovation techniques - and providing localised and enhanced results.

In our tests, each bot showcased their own unique strengths, which certainly made direct comparisons challenging.

DeepSeek's sci-fi film plot showed its creative flair that produced a more engaging and imaginative story as compared to Qwen2.5 and ChatGPT's efforts.

Unsurprisingly, the more recognized ChatGPT, unburdened by Chinese censorship constraints, supplies precise and factual reactions to concerns about Chinese current occasions, which provides it an added benefit.

Experts also weighed in on their ideas after utilizing DeepSeek and other AI apps.

"DeepSeek is at a disadvantage when it pertains to censorship constraints," kept in mind Isaac Stone Fish, creator and CEO of the research company Strategy Risks.

"When provided an option, Chinese users want the non-censored variation - simply like anyone else, so I seem like that's a piece missing out on from it."

Independent Beijing-based consultant Andy Chen Xinran said censorship would not be a dealbreaker when it pertains to AI bots, particularly for Chinese users.

"Ninety per cent of individuals using the tool are not trying to get a much deeper understanding about Xi Jinping or politically sensitive subjects. They're utilizing it for other efficient ways," Chen said.