Advertisement
Advertisement
Chinese local governments, telecommunications firms and Big Tech companies have been building new AI facilities over the past few years.
The feature accounted for over 80 per cent of recent sales made by live-streaming e-commerce firm East Buy on WeChat, local reports say.
The closed-source model bested leading models from OpenAI, Anthropic and Alibaba in some benchmark tests.
Liang’s participation at Monday’s meeting with China’s premier reflects start-up DeepSeek’s growing prominence in the AI industry.
Advertisement
The fund was established days after the US rolled out new chip export restrictions and placed more Chinese firms on its trade blacklist.
The Beijing-based tech giant has jumped on the latest trend in China’s e-commerce sector: online gifting.
Zhipu AI is among more than two dozen China and Singapore-based companies added to the US Entity List on Wednesday.
The release, which includes foundational and multimodal models, comes weeks after rival DeepSeek set a new standard with its own open-source models.
iPhone shipments surged to grab 23 per cent of the market in the final quarter of 2024, as a rebounding market saw growth start to slow.
Many Chinese firms stockpiled chips ahead of the planned roll-out of more expansive and stringent trade sanctions by Washington.
DeepSeek prefers to hire new graduates, or those early in their AI career, in line with the company’s preference for ability over experience.
01.AI denounced as ‘vicious slander’ talks that it was looking to sell its pre-training team and other assets owing to business difficulty.
Hangzhou-based DeepSeek is 2025’s ‘biggest dark horse’ in open-source large language models, Nvidia research scientist Jim Fan says.
The adjustment will position Qwen-vl-max as a direct competitor to a similar visual model ByteDance launched in the middle of December.
The shift is part of the changes to the company’s overarching algorithms that determine a series of operational procedures for its couriers.
Abundant investment and an open-source approach have pushed China to the AI forefront, but a lack of advanced chips may prove devastating.
DeepSeek’s V3 model was trained for two months at a cost of US$5.58 million, using significantly fewer computing resources than its rivals.
Fortera Capital, a private-equity firm under the Shanghai State-owned Capital Investment Co, led Stepfun’s recent series B funding round.
Chinese researchers and companies such as ByteDance made up a greater share of the 37-year-old AI conference this year, but some could not secure visas.
ByteDance’s ultra-low pricing underscores growing rivalry in the Chinese market, where Big Tech firms and start-ups have locked horns in a brutal price war.
Total funding amounted to 35 billion yuan this year, compared with 20.5 billion yuan last year, meaning investors are being more selective.
Zhipu said it doubled its revenue year on year in the 11 months to November, without providing an exact figure.
ByteDance sees AI chatbots as a slower growth opportunity and instead has prioritised AI video generators, according to local media.
The November AI model approvals is the smallest of the three batches released this year, as the market grows more saturated.
The departure of Hong Tao, who was pivotal to early monetisation efforts, comes after high-level departures at rival start-ups such as 01.AI and Moonshot AI.
The five firms that invested in Moonshot AI founder Yang Zhilin’s last venture, Recurrent AI, claim he has ‘uncleared liability’.
Some of the world’s brightest minds in AI are expected to attend this year’s Annual Conference on Neural Information Processing Systems.
The award, however, triggered speculation about whether the paper’s lead author was the ex-intern who ByteDance sued for US$1.1 million.
The company hopes to ‘better address the opportunities as well as challenges brought about by the AI 2.0 era’.
Alibaba Cloud is the latest among a slew of Chinese firms to roll out the AI models that take more time to reason through maths and programming queries.
