Baidu releases the ERNIE 5.5 Flash Preview open-source model, with 280 billion parameters and a 1 million Token context
Baidu has released ERNIE 5.5 Flash Preview, with a total parameter count of 280 billion and 14 billion parameters activated per Token, supporting a maximum context of 1 million Tokens, with outstanding performance in programming, 3D modeling, and content creation. The model is now available on Baidu AI Cloud's Qianfan platform; API input is priced at 2 yuan per million Tokens during peak hours and 1 yuan during off-peak hours, while output is 8 yuan during peak hours and 4 yuan during off-peak hours, with a limited-time 50% discount. The model weights will be released to the global open-source community on November 25.
Body · AI translation
Headlines
1. Breaking! Jensen Huang Exposed for a New Extramarital Affair with an Executive; Nvidia Board Initiates Removal, Headhunters Already Seeking His Successor
2. Mercedes-Benz CEO Ola Källenius: In China, "Cars Must Be Big, Very Big, Incredibly Big"
3. Baidu ERNIE Releases ERNIE 5.5 Flash Preview Model, Programming Capability a Major Highlight of the New Model
4. Turning Point in Control Deadlock! Nexperia CEO Says Willing to Restructure China Business: Permanent Spin-off Would Require Compensation to Wingtech Technology
5. ByteDance Pushes into the Agent Track, Plans to Launch an AI Assistant Similar to Meta Muse
6. Richard Yu Officially Announces Huawei Sub-brand "Starlight"! Stores Say Prices Won't Be Too Cheap Due to Cost Impact
7. Trip.com Launches Major Product-Research Integration, Merging Product, Front-end, Back-end and Operations, Rolling Out DRI System Company-wide
8. Apple iCloud Suffers Sudden Large-scale Outage! Mail, Photos and Other Services Unavailable Today's Headlines
Breaking! Jensen Huang Exposed for a New Extramarital Affair with an Executive; Nvidia Board Initiates Removal, Headhunters Already Seeking His Successor
November 12 news: Nvidia's stock price has collapsed once again, plunging 8% in after-hours trading. According to foreign media reports, Nvidia (NVDA) CEO Jensen Huang may have had an extramarital relationship with Kimberly Powell, a company executive who reports directly to him. This is the second personal relationship exposed by the media, following Huang's admission of an extramarital affair last week.
Citing two Nvidia insiders, foreign media reported that during an Nvidia conference held at a hotel in San Jose, California, in March this year, someone filmed a video of the Nvidia CEO and Powell behaving intimately. The foreign media added that they had not yet seen the video. After the news was published, Nvidia's stock fell more than 8% in after-hours trading. During Wednesday's regular trading session, the stock rose 5%.
It is understood that on Nvidia's company website, Powell is listed as vice president of the company's healthcare business. She joined the company in 2008 and helped Nvidia build GPUs for medical imaging instruments. According to Powell's LinkedIn profile, before joining Nvidia, she worked at Planar Systems.
Previously, Huang admitted on Sunday to an extramarital affair and a possible involvement with Alice Huang (no relation to him), an executive at data center company Megaspeed. Foreign media have just reported that after Nvidia CEO Jensen Huang was exposed for two extramarital affairs, Nvidia's board of directors is preparing to remove Huang. People familiar with the matter revealed that the board planned to hold a meeting as early as the 11th to decide his future. According to US media reports, headhunting firms have already been searching the industry for candidates to succeed Huang, including AMD CEO Lisa Su and Arm CEO Rene Haas, among others. (Bianews, Kuai Technology) Domestic News
Mercedes-Benz CEO Ola Källenius: In China, "Cars Must Be Big, Very Big, Incredibly Big"
November 11 news: According to foreign media reports, competition in the Chinese market is forcing Mercedes-Benz to adjust its product and cost strategies. Mercedes-Benz CEO Ola Källenius said product planning must be entirely centered on the preferences of Chinese consumers. First is body size: "In China, cars must be big, very big, incredibly big (original text: 'groß, sehr groß, unglaublich groß')." At the same time, Mercedes-Benz in principle needs to achieve 100% localization of its cost structure.
The first model developed according to this approach is the all-new all-electric E-Class sedan. Källenius revealed that the new car will go on sale in 2027. However, it still takes time to go from adjusting strategy to achieving results: "Decisions made today may not show results until three, four or even five years later." Mercedes-Benz can also leverage the technology-open MB.OS platform to flexibly integrate technologies from partners such as Nvidia, like assembling Lego bricks, to meet the needs of China, the United States and other markets.
Källenius believes Mercedes-Benz still has the strength to compete with Chinese automakers as a luxury brand. But he also warned that if it fails to keep up with competitors in body size, space design and digital intelligence, "it will be eliminated." Källenius also said Mercedes-Benz has launched the largest product offensive in the company's history, planning to launch or update 40 models within three years, "and these models are very popular in the market." (IT Home)
Baidu ERNIE Releases ERNIE 5.5 Flash Preview Model, Programming Capability a Major Highlight of the New Model
On November 11, Baidu's latest foundation model ERNIE 5.5 Flash Preview was officially released. ERNIE 5.5 Flash Preview has a total parameter count of 280 billion, with 14 billion parameters activated per token, supports a context length of up to 1 million tokens, and performs outstandingly in programming, 3D modeling, knowledge work, content creation and other areas.
Currently, in multiple benchmarks such as Terminal-Bench and AutomationBench, ERNIE 5.5 Flash Preview has achieved competitive scores while maintaining extremely fast decoding speed.
It is reported that ERNIE 5.5 Flash Preview has launched on Baidu AI Cloud's Qianfan platform, and developers can call the model via API. API pricing is 2 yuan per million tokens for input during peak hours and 1 yuan during off-peak hours, and 8 yuan for output during peak hours and 4 yuan during off-peak hours, with a limited-time 50% discount, giving it strong price competitiveness among models of the same parameter scale. In addition, the model weights will be officially released to the global open-source community on November 25.
A turning point in the control deadlock! Nexperia CEO says he is willing to restructure the China business: if permanently spun off, Wingtech Technology must be compensated
November 11 news, according to media reports, Stefan Tilger, interim CEO of the Netherlands' Nexperia, said in an interview during the Munich Electronics Show that although Nexperia and its China business have been operating separately for more than a year due to political factors, there is still a possibility of future reintegration between the two sides.
Tilger made clear that if the final spin-off becomes a permanent arrangement, Nexperia's parent company Wingtech Technology should receive corresponding compensation. He stressed that Wingtech Technology still enjoys the economic rights and interests of the company's shares and is only temporarily unable to exercise control, and this state should not be ignored.
However, the legal confrontation between the two sides is still escalating. On November 3, Wingtech Technology disclosed that Nexperia has filed a lawsuit with the Amsterdam District Court in the Netherlands, seeking damages of up to US$2 billion (about RMB 13.4 billion). Wingtech Technology explicitly stated that it does not recognize all the allegations and claims.
Previously, the administrator appointed by the Dutch Enterprise Chamber had proposed a plan to restructure Nexperia's equity under a "dual-circle" structure, attempting to have external investors control the international business while Wingtech controls the China business, but the plan has been rejected in writing by Wingtech Technology. At present, Nexperia China has orders and customer resources, but because the Dutch side has cut off wafer supply, production capacity has dropped to freezing point. (Kuai Technology)
November 11 news, according to foreign media reports, ByteDance plans to launch an AI assistant similar to Meta Muse within the next few weeks, consolidating its leading advantage in the field of AI video generation. ByteDance hopes to use this to accelerate its pace in catching up with top US AI labs, with the goal of narrowing the gap between the two sides in about three years.
People familiar with the matter revealed that recently, ByteDance co-founder Zhang Yiming clarified the next key priorities at an all-hands meeting for the AI team. Zhang Yiming believes that ByteDance is capable of competing with industry leaders such as Anthropic and OpenAI in the AI frontier. However, he also reminded the team that truly reaching the level of these companies will require at least until 2029 or 2030.
The new AI agent ByteDance plans to launch will play a role similar to Meta's Muse launched in September this year. Zhang Yiming has now devoted a large amount of time to behind-the-scenes work, participating in formulating ByteDance's AI strategy. In his view, one of the issues that most needs attention at present is data.
People familiar with the matter said that 43-year-old Zhang Yiming stated at the meeting that the scale of the pre-training dataset for ByteDance's Seed series of large language models is several orders of magnitude smaller than the datasets used by top US competitors. Zhang Yiming stressed that ByteDance should not overly rely on model distillation technology and should focus on long-term development. (IT Home)
Yu Chengdong officially announces Huawei sub-brand "OSTA"! Stores say prices will not be too cheap due to costs
On November 11, the official Weibo account of Huawei's all-new sub-brand "OSTA" officially went online, and Yu Chengdong simultaneously officially announced that the brand will make its first public appearance on November 20.
According to leaks, OSTA will launch two models in its first batch: the OSTA O1 with a traditional bar design, and the OSTA O1 Pro with a wide bar form factor, with configurations benchmarked against the flagship product line. The first batch of inventory is limited, and the 16GB high-configuration version will be sold only through online channels.
Regarding rumors of simultaneous offline distribution, Huawei offline store staff said they have not yet received relevant sales notices. The staff member also mentioned that due to this year's rising chip and memory prices, OSTA products are expected not to be too cheap. The market generally speculates that OSTA's positioning within the Huawei system is similar to that of the former Honor, and that the brand's primary goal at this stage is to expand market share, not to focus on profitability in the short term. (Sanyan Technology)
Ctrip launches major product-research integration, merging product, front-end, back-end and operations, and promoting the DRI system company-wide
November 11 news, according to media reports, Ctrip recently held an all-hands meeting and announced at the company level that it would begin promoting a "product-research integration" system. It is reported that this system, within the scope of implementation, uses a project-based approach to merge product, operations, front-end, back-end, algorithm, testing and other personnel, that is, "everyone is full-stack, everyone is a product manager."
It is understood that the project system is called DRI (Directly Responsible Individual), that is, the directly responsible person system. This system originated at Apple, and its core is that a project leader takes the lead, and team members collaborate quickly across functions and work together around a certain goal.
In Ctrip's specific implementation, the responsible person is not limited by function; in some teams the DRI is a former product leader, while in others it is a former operations leader or technical leader, with product and operations generally predominant; frontline members in pilot teams cover roles such as product, operations, front-end, back-end, testing and algorithms, with 1-2 people per role, so situations such as product staff evaluating technical staff's performance and technical staff reporting to operations staff may arise, and the traditional team boundaries divided by function have been broken.
In terms of authority, front-end and back-end staff within the DRI team can also raise requirements and implement them themselves; if ideas from product and operations can be realized by AI, they can also be delivered independently, as long as the DRI approves. In terms of rollout pace, Ctrip had previously run partial pilots, and currently departments under various business units such as travel, flights and the content center have responded. (Dachang Ribao)
Is it true that Tesla is recalling vehicles and paying 2,800 yuan in compensation? Tesla China responds: It is a scam, do not click the relevant links
November 11 news: A document dated November 10 has been circulating online, claiming that Tesla is implementing a recall due to "hidden manufacturing defects in the rear shock absorbers and rear springs," with compensation paid at a uniform standard of 2,800 yuan per vehicle, and the compensation funds to be distributed by the "China UnionPay Center" to owners' UnionPay accounts, Alipay or bank accounts.
In response, Tesla China's official customer service said that the text message was not officially released and is a scam, reminding owners not to click the relevant links. Customer service specifically warned that after clicking a link and entering a phone number and verification code, account deductions are very likely to occur directly. Tesla customer service also stressed that Tesla has not sent any recall text messages recently. If owners receive similar information, they should verify it through the website of the Defective Product Management Center of the State Administration for Market Regulation or by entering the Tesla App, and should not trust links attached to unfamiliar text messages.
At present, Tesla does indeed have legitimate recalls under way recently. The official website of the State Administration for Market Regulation recorded recall plans in August by Tesla (Shanghai) Co., Ltd. and Tesla Motors (Beijing) Co., Ltd., involving domestically produced and imported pure electric vehicles. Tesla China's official website also had two OTA recall announcements in November, involving respectively the rearview camera image not displaying and the door latch function. However, the handling methods for these recalls are all free software upgrades or free repairs, and no cash compensation exists. (Bianniushi)
Storage manufacturers brewing a new round of price increases? Multiple parties report that official original manufacturer prices in the fourth quarter show a slight upward trend
November 11 news: According to media reports, on the afternoon of November 11, the A-share storage chip sector rose broadly. What triggered market sentiment may have been a market rumor: some leading storage companies have orally informed channel distributors of their intention to raise prices for enterprise-grade chips, and formal price increase letters have not yet been issued, but prices on the channel side have already started moving first.
Based on feedback from multiple parties, the official prices of original manufacturers in the fourth quarter do indeed show a slight upward trend, but the models involved are still mainly enterprise-grade storage, and consumer-grade products have not shown a clear signal of simultaneous price increases. A storage module manufacturer revealed that the wafers the original manufacturers provide to module companies have actually not risen in price, and have even seen price declines, but the storage chips and module devices sold by original manufacturers and module manufacturers still show an upward price trend, especially concentrated among server manufacturers.
Previous reports showed that data indicate storage price increases are suppressing demand, and in Q3 this year China's smartphone shipments were about 58.06 million units, down 15.2% year on year. Looking ahead, storage and other component costs are expected to remain at high levels in the coming quarters, and smartphone product prices will continue to maintain an upward trend. (Shanghai Securities News)
Alibaba's all-new-generation Qwen large model Qwen4 is about to be released, and the preview version has already made its debut
November 11 news: Alibaba officially announced that the all-new-generation Qwen large model Qwen4 is about to be released: deeper knowledge understanding, more three-dimensional world perception, and continuously advancing Coding and Agentic capabilities.
Now, the Qwen4-Max preview version has made its debut on Alibaba Token Plan, where users can experience the Qwen4 model at a limited-time minimum of 10% of the price. At the same time, Qwen Office, Qoder, the Qwen App, QwenStudio and others have also integrated the new model at the first opportunity.
It is reported that the regular limited-time price of Alibaba Token Plan Personal Edition Lite/Essential starts at 39 yuan/month, and during the event new customers can enjoy an additional 50% off on top of the discount, as low as 19.5 yuan. (IT Home)
Volkswagen Anhui has no new vehicle plans for 2027 to 2028
November 11 news: According to foreign media reports, two people familiar with the matter revealed that Volkswagen's electric vehicle joint venture in China, Volkswagen Anhui, has not planned any new models for 2027 or 2028. One person familiar with the matter said that Volkswagen Anhui management previously believed that the existing product lineup was sufficient to support the brand's foothold and drive sales through 2028. In addition, Volkswagen Anhui had previously considered launching a CUV (a multi-purpose crossover model based on a sedan platform) in 2028, but the project has now been shelved.
Volkswagen China said in a statement to the media that the company is focusing on "expanding and optimizing" Volkswagen Anhui's dealer network to reach more customers and achieve sustainable development; Volkswagen declined to comment on product strategy questions.
At present, Volkswagen Anhui has launched four models, namely ID.UNYX 06, 07, 08 and 09. Previous reports said that Volkswagen Anhui delivered 1,597 units in October, with cumulative year-on-year growth of 52%. Volkswagen is continuing to advance channel adjustments in China, with some imported car stores allowed to sell Volkswagen Anhui vehicles, and the after-sales side is also being connected. (Lanjing Technology, IT Home)
MetaX and UBTECH tie up to build embodied chips; sources say joint venture Xixuan Chuangzhi raises 500 million yuan
November 11 news: According to media reports, Xixuan Chuangzhi Technology (Wuxi) Co., Ltd. recently completed 500 million yuan in angel round financing. This round was jointly led by GGV Capital and Matrix Partners China.
It is reported that this round of funding is planned mainly for the research and development of core chips and computing platforms, the building of software toolchains, the expansion of the R&D team, and joint validation with robot manufacturers and industry customers to complete mass production and delivery.
Xixuan Chuangzhi was established in May 2026, jointly initiated and founded by MetaX, one of the "four little dragons" of domestic GPUs, and UBTECH, the "first humanoid robot stock." According to the introduction, Xixuan Chuangzhi is committed to becoming a leading domestic enterprise in embodied intelligence edge computing platforms, connecting models with real tasks around chips, modules, software toolchains and engineering adaptation. (21st Century Business Herald)
Boundless Dynamics founder Zhang Yufeng responds to Li Yixian's departure: it is normal turnover, and he was not responsible for MWA pretraining
Evening news on November 11: Recently, market rumors said that Li Yixian, head of algorithm models at Boundless Dynamics, had recently resigned and joined Hello Robotaxi as head of model algorithms. In response, Boundless Dynamics founder Zhang Yufeng said: "Our algorithm models and infrastructure are still under the charge of CTO Xia Zhongpu. Yixian was 'Teacher Xia -1,' and he was not previously responsible for the pretraining of our foundation model MWA."
Public information shows that Boundless Dynamics was founded in 2025, and Zhang Yufeng came from Horizon; in June 2026, the company completed more than US$200 million in angel round financing, with JD.com-affiliated funds, Sequoia China, Linear Capital and others participating; the second-generation robot K15 has entered mass delivery, with global orders worth nearly US$100 million.
For this reason, Li Yixian's departure occurred at a stage when Boundless Dynamics was transitioning from the R&D investment period to the delivery period. The outside world once interpreted it as: the company lost core algorithm strength before the mass delivery milestone, and there were concerns that the continuity of its foundation model MWA R&D would be affected. In response, Zhang Yufeng said that there may be a misunderstanding outside, and that this is normal personnel turnover. (Sina Technology)
Some Token products are 50% off on Tmall Double 11
On November 11, it was learned that large model vendors Zhipu, MiniMax, Qianwen Office, Alibaba Cloud, Qoder and others launched 50% off activities on some products. At the Token recharge center in Tmall's AI Space Station, all three MiniMax monthly cards are 50% off, two Alibaba Cloud Token Plan monthly cards are 50% off or more, Zhipu weekly cards are 50% off or more, the first month of the Qianwen Office basic membership monthly card is 30% off or more, and the Qoder monthly card is 67% off. (Yicai)
International News
Apple iCloud suffers sudden large-scale outage! Mail, Photos and other services unavailable
On November 11, multiple Apple iCloud services suffered a sudden large-scale outage, involving Mail, Calendar, Contacts, Photos and other functions, and some users were unable to use the related services normally. Apple's official system status page showed that the outage affected a wide range, including iCloud Calendar, iCloud Contacts, iCloud Mail, iCloud Photos, iWork for iCloud, Game Center and Apple account sign-in, among other services.
In addition, Apple's Find My service also experienced abnormalities, and some users may encounter slow responses or be unable to use it normally. Affected by the outage, some users reported being unable to send and receive iCloud mail normally, and Calendar and Contacts also failed to sync normally, while others encountered problems such as abnormal Apple account sign-in. As of press time, Apple had not announced the specific cause of the outage or given a clear recovery time.
Separately, according to reports, an official Apple announcement showed that the Apple Support Community will be permanently closed on November 20, 2026. Starting immediately, the community will no longer accept new posts and will switch to read-only mode. Apple said that users can subsequently go to the official Apple Support website for product-related guidance and help. Apple also thanked community members for sharing experience over the long term and helping other users solve product usage problems. (Kuai Technology, Sanyan Technology)
OpenAI's Q3 revenue this year reaches US$10 billion for the first time, up nearly 50% quarter on quarter
November 11 news: According to foreign media reports, driven by strong demand for its latest AI products, OpenAI's third-quarter revenue reached US$10 billion (about 66.878 billion yuan) for the first time, up nearly 50% quarter on quarter and nearly three times the same period last year. A person familiar with the matter disclosed that OpenAI's adjusted operating margin improved from -150% in the previous quarter to -109% in the third quarter, narrowing its loss. In addition, OpenAI said on a conference call that as of the end of September, its annualized revenue scale was about US$50 billion (about 334.388 billion yuan).
This year, OpenAI had lagged behind long-time rival Anthropic on some metrics, but with the improved competitiveness of its new-generation AI models, sharp product price cuts, and a further shift in business focus toward AI agents that help customers automate work, OpenAI has regained growth momentum. Anthropic's revenue growth has begun to slow, while OpenAI's business growth shows signs of accelerating.
Anthropic, which is about to list in the US stock market, had third-quarter revenue of about US$17 billion (about 113.692 billion yuan), up about 47% quarter on quarter, a marked slowdown from the more than 140% increase in the second quarter. However, the two companies' past revenue statistics are not on the same basis. OpenAI has postponed its listing plan to next year and is currently completing a new round of financing at a valuation of US$1.25 trillion (about 8.36 trillion yuan). Anthropic plans to list later this month, with a target valuation of US$1.4 trillion (about 9.36 trillion yuan). (IT Home)
Report: SpaceX explores "de-NVIDIA-fication" at the server rack level: buy only chips, procure more supporting hardware itself
November 11: Semiconductor research firm SemiAnalysis disclosed on X that SpaceX is doing something previously attempted only by hyperscale cloud providers: keeping NVIDIA's chips but rebuilding the racks around the chips itself. The internal codename for this self-developed rack is Kessel.
SemiAnalysis said SpaceX plans to purchase chip packages directly from NVIDIA and procure the remaining components itself, thereby skipping NVIDIA's L6 compute boards. The firm wrote: "This also changes who buys which parts: SpaceX plans to buy chip packages from NVIDIA and procure the rest directly, bypassing NVIDIA's L6 compute boards. This will give SpaceX more control over supplier selection and component procurement." Skipping L6 means SpaceX has gained design and procurement rights at the board level—PCB, connectors, power supply and cooling shift from "specified by NVIDIA" to "chosen by SpaceX." (Bianews)
Google launches Gemini 4 Carbon project globally
November 11 news: Google announced on Tuesday that its Gemini 4 Carbon detector will be opened to more users worldwide, including developers, scientists, enterprises and consumers.
The company calls it its "most powerful model," with enhanced cybersecurity capabilities and the ability to handle long-running agentic workflows spanning complex knowledge tasks, software engineering and scientific reasoning.
Specifically, Carbon helped its first batch of users discover zero-day vulnerabilities in a short period of time. Gemini 4 Carbon is now available in Google AI Studio and Gemini for Business. (Sanyan Technology)