Alibaba rolls out latest version of its large language model to meet robust AI demand
The logo of the Alibaba office building is seen in the Huangpu District in Shanghai, June 16, 2023.
Costfoto | Nurphoto | Getty Images
Alibaba Cloud said on Thursday it released the latest version of its large language model after more than 90,000 deployments by companies.
Jingren Zhou, chief technology officer of Alibaba Cloud, said in a statement the firm has seen “many creative applications of the models from across the industries,” including consumer electronics and gaming.
“We look forward to collaborating with our customers and developers in seizing the immense
growth opportunities presented by the latest surge in the generative AI development,” said Zhou.
Alibaba Cloud said the latest version of its Tongyi Qianwen model, Qwen2.5, possesses “remarkable advancements in reasoning, code comprehension, and textual understanding compared to its predecessor Qwen2.0.”
Large language models power artificial intelligence applications like OpenAI’s ChatGPT. They are trained on vast amounts of data to generate humanlike responses to user prompts.
The latest Qwen model fares better than OpenAI’s GPT-4 model in language and creation capabilities, but fell short in other categories like knowledge, reasoning and math, according to a March analysis by large language model evaluation platform OpenCompass.
Alibaba released Tongyi Qianwen in April 2023 after ChatGPT took the world by storm following its November 2022 launch. An upgraded version was released in October with improved capabilities in understanding complex instructions, copywriting, reasoning, memorizing, among others.
Alibaba Cloud said more than 2.2 million corporate users have accessed Qwen-powered AI services such as DingTalk – Alibaba’s answer to Slack.
The firm also said it has launched a series of new Qwen models to the open-source community and upgraded Model Studio, its generative AI platform, with new AI development resources.
Other Chinese tech giants such as Baidu and Tencent have released similar chatbots and AI models amid explosive demand for generative AI. Baidu in April said its Ernie bot has exceeded 200 million users after obtaining the green light for public use in August.
Generative AI is accelerating development of humanoid robots in China where these robots could help with factory work or labor.
Technology
‘Robotaxi has reached a tipping point’: Baidu, Nvidia leaders see momentum as competition rises
Chinese tech company Baidu announced Monday it can sell some robotaxi rides without any human staff in the vehicles.
Baidu
BEIJING — Chinese robotaxi companies are expanding abroad at a faster clip than U.S. rivals Waymo and Tesla — at a time when industry leaders say autonomous driving is finally near an inflection point.
“I think robotaxi has reached a tipping point, both here in China and in the U.S.,” Baidu CEO Robin Li said Tuesday on an earnings call, according to a FactSet transcript.
“There are enough people who have [had the] chance to experience driverless rides, and the word of mouth has created positive social media feedback,” he said, noting that the wider public exposure could speed up regulatory approval.
His comments echoed similar notes of optimism in the last few weeks from Nvidia CEO Jensen Huang and Xpeng Co-President Brian Gu — who reversed his previously cautious stance after faster-than-anticipated tech advances. Xpeng is launching robotaxis in the southern Chinese city of Guangzhou next year.
It’s a global market with significant growth potential, likely worth more than $25 billion by 2030, according to Goldman Sachs’ estimates in May.
To seize that opportunity, Chinese companies are aggressively expanding overseas and claim they are close to making robotaxis a viable business, rather than simply burning cash to grab market share.
In the last 18 months, Baidu, Pony.ai and WeRide landed partnerships with Uber that allow users of the ride-hailing app to order a robotaxi in specific locations, starting in the Middle East.
Such tie-ups “will be critical to success” as they enable robotaxi companies to operate more efficiently and reach profitability more quickly, said Counterpoint Senior Analyst Murtuza Ali.
Once we can generate profit for every single car in a second-tier city [like Wuhan] in mainland China, we can generate profits in lots of cities across the world.
Halton Niu
General manager for Apollo Go’s overseas business
Expanding on experience at home
Baidu says that since late last year, its Apollo Go robotaxi unit has reached per-vehicle profitability in Wuhan, where the company has operated over 1,000 vehicles in its largest deployment in China.
That means ridership is enough to offset a Wuhan taxi fare that’s 30% cheaper than in Beijing or Shanghai, and far below prices in the U.S. or Europe. Besides developing autonomous driving systems, Baidu has also produced electrically-powered robotaxi vehicles — without relying on a third-party manufacturer — that are 50% cheaper.
“Once we can generate profit for every single car in a second-tier city [like Wuhan] in mainland China, we can generate profits in lots of cities across the world,” Halton Niu, general manager for Apollo Go’s overseas business, told CNBC.
“Scale matters,” he said. “If you only deploy, for example, 100 to 200 cars in a single city, if you only cover a small area of the city, you can never become profitable.”
How U.S. rivals stack up
Scale remains the dividing line. In the U.S., Alphabet-owned Waymo operates more than 2,500 vehicles and is expanding rapidly from major cities in California to Texas and Florida, with plans to enter London next year, following its first overseas venture in Tokyo.
Tesla sells its electric cars in China, and reportedly showed off its Cybercab in Shanghai this month. But it began testing its robotaxis in Texas only in June, and this week obtained a permit to operate in Arizona.
Amazon’s Zoox is also ramping up its expansion in the U.S., but has not released overseas plans.
The three companies have not disclosed plans to break even on their robotaxis.
Baidu Apollo Go’s Niu did not rule out an expansion into the U.S. But for now, the robotaxi operator plans to enter Europe with trials in parts of Switzerland next month, following their expansion in the Middle East this year.
Abu Dhabi last week gave Apollo Go a permit to charge fares to the public for fully driverless robotaxi rides, which are operated locally under the AutoGo brand, eight months after local trials began in parts of the city.
But Chinese startup WeRide said it received a similar permit on Oct. 31 to charge fares for its fully driverless robotaxi rides in Abu Dhabi, and claimed that removing human staff from the cars would allow it to make a profit on each vehicle.
That puts Pony.ai furthest from profitability among the three major Chinese robotaxi operators. Its CFO Leo Haojun Wang told The Wall Street Journal in mid-September that the company aimed to make a profit on each car by the end of this year or early next year.
Pony.ai plans to launch a fully autonomous commercial robotaxi business in Dubai in 2026, after receiving a testing permit in late September. The company plans to roll out in Europe in the coming months and has also outlined an expansion into Singapore.
Pony.ai and WeRide are set to release quarterly earnings early next week.
“Currently, companies like Waymo, Baidu, WeRide and Pony.ai are leading in terms of fleet size, which positions them advantageously in the race for profitability,” said Yuqian Ding, head of China Autos Research at HSBC.
Scale and safety
Fleet size is becoming a competitive marker. Pony.ai reportedly said it plans to release 1,000 robotaxis in the Middle East by 2028, while WeRide aims to operate a fleet of 1,000 robotaxis in the region by the end of next year.
Niu said Apollo Go operates around 100 robotaxis in Abu Dhabi and Dubai, and plans to double its vehicle fleet in the next few months.
“Apollo Go has had a head start with significantly more test rides than the other two,” Kai Wang, Asia equity market strategist at Morningstar, said in an email. “The more testing and data you can collect from trips taken, the more likely the AI sensors are able to recognize the objects on the road, which means better safety as well.”
He cautioned that despite some initial progress, the robotaxi race remains uncertain as “no one has truly had mass adoption for their vehicles.”
Subscribe now
Coverage remains limited. Even in China, robotaxis are only allowed to operate in selected zones, though Pony.ai recently became the first to win regulatory approval to operate its robotaxis across all of Shenzhen, dubbed China’s Silicon Valley. In Beijing, self-driving taxis are mostly limited to a suburb called Yizhuang.
Anecdotally, CNBC tests have found Pony.ai offered a smoother ride than Apollo Go, which was prone to hard braking.
As for safety — which is critical for regulatory approval — none of the six operators has reported fatalities or major injuries caused by the robotaxis so far. But Apollo Go and Waymo have begun advertising low airbag deployment rates.
Even if that’s not enough to convince regulators worldwide, Beijing is expected to ramp up support at home.
HSBC’s Ding predicts the number of robotaxis on China’s roads could multiply from a few thousand to tens of thousands between the end of this year and 2026, a shift that would give operators more proof that their model works.
Jensen Huang, chief executive officer of Nvidia Corp., during the US-Saudi Investment Forum at the Kennedy Center in Washington, DC, US, on Wednesday, Nov. 19, 2025.
Stefani Reynolds | Bloomberg | Getty Images
In the weeks leading up to Nvidia’s third-quarter earnings report, investors debated whether the markets were in an AI bubble, fretting over the massive sums being committed to building data centers and whether they could provide a long-term return on investment.
During Wednesday’s earnings call with analysts, Nvidia CEO Jensen Huang began his comments by rejecting that premise.
“There’s been a lot of talk about an AI bubble,” Huang said. “From our vantage point we see something very different.”
In many respects, Huang’s remarks are to be expected. He’s leading the company at the heart of the artificial intelligence boom, and has built its market cap to $4.5 trillion because of soaring demand for Nvidia’s graphics processing units.
Huang’s smackdown of bubble talk matters because Nvidia counts every major cloud provider — Amazon, Microsoft, Google, and Oracle — as a customer. Most of the major AI model developers, including OpenAI, Anthropic, xAI and Meta, are also big buyers of Nvidia GPUs.
Huang has deep visibility into the market, and on the call he offered a three-pronged argument for why we’re not in a bubble.
First, he said that areas like data processing, ad recommendations, search systems, and engineering, are turning to GPUs because they need the AI. That means older computing infrastructure based around the central processor will transition to new systems running on Nvidia’s chips.
Second, Huang said, AI isn’t just being integrated into current applications, but it will enable entirely new ones.
Finally, according to Huang, “agentic AI,” or applications that can run without significant input from the user, will be able to reason and plan, and will require even more computing power.
In making the case of Nvidia, Huang said it’s the only company that can address the three use cases.
“As you consider infrastructure investments, consider these three fundamental dynamics,” Huang said. “Each will contribute to infrastructure growth in the coming years.”
Reversing the slide
In its earnings release, Nvidia reported revenue and profit that sailed past estimates and issued better-than-expected guidance. Last month, Huang provided a $500 billion forecast for sales of the company’s AI chips over calendar 2025 and 2026.
The company said on Wednesday that its order backlog didn’t even include a few recent deals, like an agreement with Anthropic that was announced this week or the expansion of a deal with Saudi Arabia.
“The number will grow,” CFO Colette Kress said on the call, saying the company was on track to hit the forecast.
Prior to Wednesday’s results, Nvidia shares were down about 8% this month. Other stocks tied to the AI have gotten hit even harder, with CoreWeave plunging 44% in November, Oracle dropping 14% and Palantir falling 17%.
Some of the worry on Wall Street has been tied to the debt that certain companies have used to finance their infrastructure buildouts.
“Our customers’ financing is up to them,” Huang said.
Specific to Nvidia, investors have raised concerns in recent weeks about how much of the company’s sales were going to a small number of hyperscalers.
Last month, Microsoft, Meta, Amazon and Alphabet all lifted their forecasts for capital expenditures due to their AI buildouts, and now collectively expect to spend more than $380 billion this year.
Huang said that even without a new business model, Nvidia’s chips boost hyperscaler revenue, because they power recommendation systems for short videos, books, and ads.
People will soon start appreciating what’s happening underneath the surface of the AI boom, Huang said, versus “the simplistic view of what’s happening to capex and investment.”
WATCH: Nvidia posts Q3 beat
-
Sports2 years agoStory injured on diving stop, exits Red Sox game
-
Sports3 years ago‘Storybook stuff’: Inside the night Bryce Harper sent the Phillies to the World Series
-
Sports2 years agoGame 1 of WS least-watched in recorded history
-
Sports3 years agoButton battles heat exhaustion in NASCAR debut
-
Sports3 years agoMLB Rank 2023: Ranking baseball’s top 100 players
-
Sports4 years ago
Team Europe easily wins 4th straight Laver Cup
-
Environment2 years agoJapan and South Korea have a lot at stake in a free and open South China Sea
-
Environment1 year agoHere are the best electric bikes you can buy at every price level in October 2024