Chinese company Zhipu AI has trained image generation model entirely on Huawei processors, demonstrating that Chinese firms can build competitive AI systems without access to advanced Western chips.

The model, released on Tuesday, marks the first time a state-of-the-art multimodal model completed its full training cycle on Chinese-made chips, Zhipu said in a statement. The Beijing-based company trained the model on Huawei’s Ascend Atlas 800T A2 devices using the MindSpore AI framework, completing the entire pipeline from data preprocessing through large-scale training without relying on Western hardware.

The achievement carries strategic significance for Zhipu, which the US Commerce Department last year added to a list of entities acting contrary to US national security or foreign policy interests over its alleged ties to China’s military. The designation effectively cut the company off from Nvidia’s H100 and A100 GPUs, which have become standard for training advanced AI models, forcing Chinese firms to develop alternatives around domestic chip architectures.

Followign that listing, Zhipu began collaborating with Huawei on GLM-Image. Huawei’s Ascend processors have become the primary alternative for Chinese AI companies restricted from purchasing Nvidia’s hardware. The model’s successful training on Ascend chips provides a data point that Chinese firms can develop competitive AI systems despite restricted access to Western chips.

“This proves the feasibility of training high-performance multimodal generative models on a domestically developed full-stack computing platform,” Zhipu’s statement added.

Zhipu has made GLM-Image available through an API for 0.1 yuan (approximately $0.014) per generated image. The company released the model weights on GitHub, Hugging Face, and ModelScope Community for independent deployment.

The pricing positions GLM-Image as a cost-effective option for enterprises generating marketing materials, presentations, and other text-heavy visual content at scale.

Technical approach and benchmark performance

GLM-Image employs a hybrid architecture combining a 9-billion-parameter autoregressive model with a 7-billion-parameter diffusion decoder, according to Zhipu’s technical report. The autoregressive component handles instruction understanding and overall image composition, while the diffusion decoder focuses on rendering fine details and accurate text.

The architecture addresses challenges in generating knowledge-intensive visual content where both semantic understanding and precise text rendering matter, such as presentation slides, infographics, and commercial posters.

On the CVTG-2K benchmark, which measures accuracy in placing text across multiple image locations, GLM-Image achieved a Word Accuracy score of 0.9116, ranking first among open-source models. The model also led the LongText-Bench test for rendering extended text passages, scoring 0.952 for English and 0.979 for Chinese across eight scenarios including signs, posters, and dialog boxes.

The model natively supports multiple resolutions from 1024×1024 to 2048×2048 pixels without requiring retraining, the report added.

Hardware optimization strategy

Training GLM-Image on Ascend hardware required Zhipu to develop custom optimization techniques for Huawei’s chip architecture. The company built a training suite that implements dynamic graph multi-level pipelined deployment, enabling different stages of the training process to run concurrently and reducing bottlenecks.

Zhipu also created high-performance fusion operators compatible with Ascend’s architecture and employed multi-stream parallelism to overlap communication and computation operations during distributed training. These optimizations aim to extract maximum performance from hardware that operates differently from the Nvidia GPUs most AI frameworks target by default.

The technical approach validates that competitive AI models can be trained on China’s domestic chip ecosystem, though at what cost in development time and engineering effort remains unclear.

Zhipu did not say how many processors or how long it took to train its model, nor how the requirements compared to equivalent Nvidia-based systems.

Implications for global AI development

For multinational enterprises operating in China, GLM-Image’s training on domestic hardware provides evidence that Chinese AI infrastructure can support state-of-the-art model development. Companies with Chinese operations may need to evaluate whether to develop strategies around platforms like Huawei’s Ascend and frameworks like MindSpore.

The release comes as Chinese companies invest in domestic AI infrastructure alternatives. Whether export controls will slow or accelerate the development of parallel AI ecosystems remains a subject of policy debate.

This article first appeared on Infoworld.

Source:: Computer World

By Ana-Maria Stanciuc Belgian cybersecurity startup Aikido Security has closed a $60 million Series B funding round at a $1 billion valuation, marking a rare unicorn milestone for a European security company and highlighting the accelerating interest in developer-centric security platforms. The round was led by DST Global, a tech-oriented investment firm with a track record of backing major technology companies, and included participation from PSG Equity, Notion Capital, and Singular. Founded in 2022, Aikido has reached a billion-dollar valuation in just three years, a pace few cybersecurity firms attain, and according to company sources, the fastest in Europe. The new capital will…This story continues at The Next Web

Source:: The Next Web

By Paulo Vargas Samsung XR smart glasses may launch in two versions. A leak cites model numbers SM-O200P and SM-O200J, plus a 12MP autofocus camera, Wi-Fi and Bluetooth, and transition lenses on one variant.
The post Samsung XR smart glasses leak says you may see two versions appeared first on Digital Trends.

Source:: Digital Trends

Apple on Tuesday confirmed that it’s working with Google Gemini to build AI Foundation Models used across its platforms. As the joint statement explained it: “Apple and Google have entered into a multi-year collaboration under which the next generation of Apple Foundation Models will be based on Google’s Gemini models and cloud technology. These models will help power future Apple Intelligence features, including a more personalized Siri coming this year.

“After careful evaluation, Apple determined that Google’s Al technology provides the most capable foundation for Apple Foundation Models and is excited about the innovative new experiences it will unlock for Apple users. Apple Intelligence will continue to run on Apple devices and Private Cloud Compute, while maintaining Apple’s industry-leading privacy standards.”

What is the partnership structure?

The announcement confirms that Apple and Google have entered into a multi-year collaboration partnership in which the next generation of Apple Foundation Models will be based on Google’s Gemini models.

Is it an exclusive deal?

The arrangement allows Apple to work with other AI providers, including OpenAI, if it wants.

What about the money?

We do not know the financial terms of the deal. Bloomberg at one point claimed Apple intended to pay around $1 billion a year to use Gemini, but the actual terms weren’t disclosed, nor are they likely to be. But both companies can already anticipate additional scrutiny on their financial statements in the coming months.

What is the deployment model?

Apple’s plan is to use its own self-customized version of Gemini, tweaked to make sure queries are handled in Apple’s preferred fashion. Apple Insider reports the absence of any mention of Google or Gemini in the UI. Essentially, Apple will use Gemini to form the base of Apple Foundation Models it will make for itself. Apple can also ask Google to tweak aspects of how the Gemini models work.

What price privacy?

Apple cares about privacy. That is why the AI features will run either on-device or using Apple’s Private Cloud Compute, not on Google’s servers. It will be possible to use third-party cloud services for complex problems, if required. But the use of Apple’s own servers means Google won’t have direct access to user data.

What will happen as a result?

With Gemini as its foundation, Apple can now deliver:

A major Siri overhaul, rolling out this spring.

Even more evolved on-device contextual understanding that will follow in months, as Apple Intelligence becomes able to figure out data, such as identifying your relatives, on your behalf.

A Siri that’s better at conversational responses and also more likely to at least try to find the correct response, rather than saying it doesn’t understand.

The ability to create documents, and eventually remember past conversations and make proactive suggestions based on information from your apps

It is important to note that as Apple builds future Foundation Models on the foundation of Gemini, Apple developers will likely also gain access to using those models within their apps. 

Why Gemini?

Apple is relying on Gemini to form the base model for its own AI development. It will then tweak and train those models to create its own agent. Max Weinbach has written an excellent review of how this could work. He estimates that Apple is effectively covering the cost of Google’s future models within the deal.

Why did this happen?

We’ve had months of reporting about how and why Apple seemingly missed a trick with AI, and at least one technical problem has been publicly disclosed. Craig Federighi, Apple’s vice president for software, told staffers that Apple had intended to merge its existing automation systems with generative AI. “We initially wanted to do a hybrid architecture, but we realized that approach wasn’t going to get us to Apple quality,” he said.

We’ve also heard about internal conflicts, poor leadership, and unexpected challenges in the work. This may not matter now as Apple leans into Gemini.

What analysts say

“For Apple, partnering, rather than building an end-to-end AI proprietary model stack, could compress time-to-market and reduce execution risk by leveraging mature, already-deployed technology.” — Anisha Bhatia, senior technology analyst at GlobalData.

“For the people asking if the Apple-Google deal means there is no differentiation, think about it in F1 terms: multiple teams run the same engine, yet deliver vastly different results based on design and setup. Same here.” — Creative Strategies analyst Carolina Milanesi.

“Apple’s decision to use Google’s Gemini models for Siri shifts OpenAI into a more supporting role, with ChatGPT remaining positioned for complex, opt-in queries rather than the default intelligence layer.” — Parth Talsania, CEO of Equisights Research.

What is the rollout schedule?

Apple will introduce some new features this spring, aiming to keep to earlier promises. More sophisticated features, including proactive AI, will likely be announced at WWDC for introduction later this year, The Information reports.

What next?

Ask Siri in spring.

You can follow me on social media! Join me on BlueSky,  LinkedIn, Mastodon, and MeWe. 

Source:: Computer World

By Nadeem Sarwar The RayNeo X3 Pro smart glasses are built atop Android foundations and heavily rely on Gemini, serving it all atop an invisible screen and “normal” form factor. It’s only plagued by the curse of a first generation product.
The post I finally found AI smart glasses that offer a true glimpse of the future appeared first on Digital Trends.

Source:: Digital Trends

By Alexandru Stan A private network for founders and executives, limited to 1,000 members The Next Web today announces the launch of TNW Council, a private membership community created for founders and executives leading technology companies. The Council is limited to 1,000 members globally. Each member is selected through an application and review process. The Council is designed to bring together experienced leaders who want peer conversations, meaningful connections, and visibility through editorial and event participation. Members gain editorial participation opportunities across the platform, access to curated expert panels, and verified executive profiles. They receive priority access to conferences, invitations to private gatherings…This story continues at The Next Web

Source:: The Next Web

By Deepti Pathak Samsung fans are already looking ahead to the Galaxy S26 Ultra, which is expected to arrive…
The post Samsung Galaxy S26 Ultra Rumours: New Design, Faster Charging, and a Delayed March Launch appeared first on Fossbytes.

Source:: Fossbytes

By Moinak Pal A new near-planar OLED light extraction method could deliver brighter, more efficient displays for smartphones, TVs, and future flexible devices.
The post Scientists are teaching OLED screens how to shine smarter appeared first on Digital Trends.

Source:: Digital Trends

By Shikhar Mehrotra A hidden Google app suggests the company’s smart glasses won’t scream for attention; instead, they’ll listen, pause, and politely step aside when real life takes over.
The post Google app leak gives us a glimpse of life with smart glasses and meaningful controls appeared first on Digital Trends.

Source:: Digital Trends

By Omair Khaliq Sultan If you like the idea of health tracking but you don’t want a smartwatch on your wrist 24/7, a smart ring is the most “set it and forget it” option. The Samsung Galaxy Ring is down to $299.99, saving you $100 off the $399.99 compared value. It’s a meaningful discount on a newer category of […] The post The Samsung Galaxy Ring is $100 off right now at $299.99 appeared first on Digital Trends.

Source:: Digital Trends

Slack has launched the revamped version of Slackbot, with generative AI (genAI) features that let workers more easily find information and create documents. 

The feature is generally available for certain paid Slack plans as of today, the company announced Tuesday. Down the road, it will coordinate other AI agents to complete tasks in other apps.

“The new Slackbot release addresses a common AI challenge: generating accurate responses that are informed by relevant context,” said Wayne Kurtzman, research vice president at IDC. “Since Slack is often integrated across a large number of enterprise applications, it a logical choice to create an enterprise search that lives within the flow of work.”

There are already numerous AI assistants available to office workers, from general purpose assistants such as OpenAI’s ChatGPT and Anthropic’s Claude to the AI tools built into productivity apps sold by Google, Microsoft, Notion, Zoom, and others. 

Slackbot’s strength is that it’s built natively into Slack’s collaboration app, said Rob Seaman, chief product officer and interim CEO at Slack. This means it immediately has knowledge of a user’s work life, with access to conversation and files, as well as enterprise data from connected sources. 

“The advantages are proximity and context,” Seaman said in an interview. “It knows what you’re working on, and it knows what your company’s priorities are because it can access what you have access to within Slack.” 

Slackbot has been around since Slack launched more than a decade ago. Slack users will likely have received notifications via the bot, or used it for simple tasks such as setting reminders. The previous version relied on rule-based automation, while the updated assistant is powered by a large language model (LLM). (Slack declined to say which LLM it uses; CNBC and Venturebeat reported that it relies on  Claude.)

Slack users can converse with Slackbot in natural language as they do with most AI assistants. One of the primary use cases is search: Slackbot can query all data a user is permitted to access across Slack and Salesforce, as well as connected enterprise sources such as Google Drive and OneDrive. This can help users track down files, or get a summary of the progress of a particular project, for instance. 

Slackbot can also generate Slack canvas documents based on user prompts, and check calendars to help set up meetings between coworkers.

Salesforce, Slack’s parent company, trialed the AI assistant internally prior to launch, with 25,000 weekly active users in just a few weeks “through word of mouth,” said Seaman. “We didn’t do any in-product nudging.”

Seaman highlighted some of the ways the AI assistant has been used by the Slack product engineering team. Before a recent all-hands meeting, for instance, he asked Slackbot check the meeting deck and provide guidance on how to pronounce the names of more than 60 new hires. The AI assistant has also been used to collate product development feedback from a Slack channel. 

“In seconds, I have a summarized view of the feedback — positive and negative — which would have taken a product manager hours or maybe a week historically,” he said. Slackbot then turned the feedback into a product brief and a sprint plan, creating canvas documents with information for the engineering team, he said. And sales teams at the company used Slackbot to get a quick rundown of clients they haven’t spoken to yet in relation to a particular product. 

Seaman also described plans to enable Slackbot to interact with other AI agents. 

There are already many examples of customers deploying agents into Slack, he said, whether custom-built Agentforce agents or the likes of Atlassian, Cursor, Claude Code, and OpenAI Codex. “Users shouldn’t have to think about what agent to go to and for what task,” he said, stressing the company wants Slackbot to “act as a concierge and make hand-offs to the appropriate task or tool or agent in the moment.” That will mean Slackbot acting as a model context protocol (MCP) client as well as server, he said, though the “handoff” will be built into the Slack UI. 

“We think of it as a super agent that’s going to help you connect with these other agents … and ideally they speak to each other,” said Seaman.

These features are still under development and not currently available. 

Other planned capabilities include the ability for Slackbot to access a user’s screen (opt-in required) to view information such as a canvas document they are working on. 

Slackbot is now generally available to customers on Business+ and Enterprise+ plans. The rollout follows pricing changes last year that eliminated separate AI add-ons and raised the price of the Business+ tier to $15 per user per month.

Source:: Computer World

Apple is going full throttle in the run-up to its Q1 FY26 fiscal call at the end of this month, boosting investor confidence with its AI partnership with Google and following that up with an announcement to inject even more growth into its all-important services division.

Introducing Apple Creator Studio

For decades, Apple has published two of the industry’s leading products for creative professionals, Final Cut Pro and Logic Pro. It has accompanied these with powerful tools such as Motion, Compressor, and Main Stage, and useful productivity apps including Pages, Keynote, and Numbers. (Apple also acquired Pixelmator Pro in 2024.)

All these apps have now been combined into one subscription-based product Apple calls Creator Studio. For just $12.99 per month, Apple customers can subscribe to and use all of these creative apps, including advanced tools not available in the free versions of some of the included applications. The apps are all available for Macs. most are available for the iPad, and some (such as Keynote or Pages) are also for iPhone users.

Apple continues to sell the Mac versions of Final Cut Pro, Pixelmator Pro, Logic Pro, Motion, Compressor, and MainStage as one-time purchases on the Mac App Store, while Keynote, Numbers, and Pages are also still available for free. And Apple’s powerful image editing application, Pixelmator Pro, is available for the iPad — and compatible with Apple Pencil.

One thing that should be clear is that with its industry leading creative tools, Creator Studio represents an immediate challenge to other far more expensive imaging, video, and audio suites on the market.

What you get in Apple Creator Studio 

In brief, the suite offers the following:

Final Cut Pro, Logic Pro, and Pixelmator Pro on Macs and iPads.

Motion, Compressor, and MainStage on the Mac.

Intelligent features and premium content for Keynote, Pages, and Numbers, with Freeform support to follow.

While Mac users can continue to purchase one-time versions of all the apps, iPad users will need a subscription to access the suite. Apple stresses that Pages, Numbers, and Keynote will “remain free for all users to create, edit, and collaborate with others” and will also continue to receive updates.

“Apple Creator Studio is a great value that enables creators of all types to pursue their craft and grow their skills by providing easy access to the most powerful and intuitive tools for video editing, music making, creative imaging, and visual productivity — all levelled up with advanced intelligent tools to augment and accelerate workflows,” Eddy Cue, Apple’s senior vice president of internet software and services, said in a statement. 

“There’s never been a more flexible and accessible way to get started with such a powerful collection of creative apps for professionals, emerging artists, entrepreneurs, students, and educators to do their best work and explore their creative interests from start to finish,” he said.

What’s the approach?

So, if some of the apps are available free or as standalone purchases, why switch to a subscription? Apple’s approach is two-fold: 

First, given the prevalence of piracy when it comes to powerful creative apps, making software available at prices people can more easily afford makes sense; $12.99 for leading tools for audio, video, and photography seems like an approachable price, which should go some way to preventing people from downloading malware invested versions of these apps from various dodgy file repositories. Security firms, including Jamf, have identified many instances in which pirated versions of Final Cut have been infested with malware. 

Second, it looks as if Apple will field new tools and features for the subscription-based apps first. Keynote, Pages, Numbers, and Freeform users will be able to access new premium content and intelligent features across Macs, iPads, and iPhones, the company said.

With 2.5 million people already using Final Cut Pro, Apple must surely hope the suite will be a sufficiently compelling proposition to tempt them to switch to a subscription.

Now with Apple Intelligence (and friends)

Hot on the heels of news of its AI partnership with Google Gemini, it should be of no surprise that Apple is also weaving a series of intelligent new features into the Creator Studio apps. It is also easy to imagine how Apple will be able to augment all of its creative apps with smart features down the road.

Among the smart features within the current Creator Studio:

You can search for specific soundbites across hours of video footage by typing the required phrase in Final Cut.

You find specific clips and actions using Visual Search.

Beat Detection uses AI to instantly analyze any music track to display on a grid in Final Cut to help align cuts to the music.

You get a new AI driven Synth Player in Logic Pro.

Chord ID can turn any audio or MIDI recording into a ready-to-use chord progression.

You can search for specific or similar loops in Logic Pro tracks using written prompts.

Keynote, Pages, and Numbers gain access to the Content Hub, a repository of high-quality images, graphics, and illustrations, along with new premium templates and themes. The three apps also gain advanced image creation and editing tools to craft high-quality images from text using generative models from OpenAI.

Keynote subscribers will be able to use a smattering of in-development AI features, such as the ability to generate a first draft of a presentation from a text outline.

Apple’s services segment generated more than $100 billion in revenue in 2025. Developing this side of the multinational enterprise has been a big success for the company, particularly as the margins it generates on services are very high (75%). The decision to introduce Apple Creator Studio as a service can only drive these numbers higher.

The suite will be available beginning on Jan 28. In addition to the $12.99 per month price, users can opt to pay $129 a year. 

Please follow me on LinkedIn, Mastodon, or join me in the AppleHolic’s bar & grill group on MeWe.

Source:: Computer World

By Ana-Maria Stanciuc For years, online shopping followed a fixed path. Search bar, product grid, filters, checkout. The interface barely changed, even as everything else on the internet did. That path is now breaking. Monday 12 January, 2026, JD Sports Fashion plc (JD Group) became one of the first large retailers to let US customers search and buy sneakers directly inside AI chatbots, including ChatGPT, Microsoft Copilot, and Google Gemini. No website visit. No app. Just a conversation that ends with a payment. Basically, a customer can chat with an AI bot, say “add these shoes to my cart,” and complete a purchase in…This story continues at The Next Web

Source:: The Next Web

By Pranob Mehrotra Anthropic’s latest tool lets Claude access local files to handle mundane tasks on your behalf.
The post Anthropic’s Cowork turns Claude into your hands-on digital teammate appeared first on Digital Trends.

Source:: Digital Trends

The Guardian has published a number of articles highlighting the risks of asking health-related questions to chatbots or other AI tools.

For example, Google’s AI overviews provide directly dangerous answers about liver values, which can lead people with serious liver diseases to believe that they are healthy.

Following the review, Google has decided to remove some of the criticized AI summaries. However, there is still a risk that they will appear if the user changes to a different wording.

“There are still far too many examples where Google’s AI summaries give people incorrect health information,” said Sue Farrington of the Patient Information Forum in a comment to The Guardian.

This article originally appeared on Computer Sweden.

Related reading:

Common health questions to ask Chat GPT

AI chatbots ditch medical disclaimers, putting users at risk, study warns

Will AI help doctors decide whether you live or die?

GenAI is already transforming the healthcare industry

Source:: Computer World

Google has introduced several new tools and standards to help retailers with AI-driven commerce, where digital agents handle purchases for users.

The biggest news is the Universal Commerce Protocol (UCP). An open standard that enables different AI agents, payment systems, and stores to collaborate through a common language.

The protocol has been developed in collaboration with players such as Shopify, Etsy, Walmart, and Target, and is supported by over 20 other companies, including Mastercard, Visa, and Zalando.

UCP will first be used for direct checkout features in Google’s AI mode in Search and in the Gemini app, where users can complete purchases with Google Pay or PayPal without leaving Google’s interface.

This article originally appeared on Computer Sweden.

More Google news:

Google may soon let users change their Gmail address

Google is working on Nano Banana 2 Flash to generate images at lightning speed

Google’s new AI tool Disco can turn tabs into apps

Source:: Computer World

By Alexandru Stan For the past few years, artificial intelligence has been discussed almost exclusively in terms of models. Bigger models, faster models, smarter models. More recently, the focus shifted to agents, systems capable of planning, reasoning, and acting autonomously. Yet the real leap in usefulness does not happen at the model level, nor at the agent level. It happens one layer above, at the level of Skills. If models represent intelligence and agents represent coordination, Skills are where AI becomes operational and valuable in the real world. A Skill is not a prompt. It is not a chatbot. And not an agent.…This story continues at The Next Web

Source:: The Next Web

By Pranob Mehrotra Anthropic’s new Claude for Healthcare will help you understand health and fitness metrics from wearables like the Apple Watch.
The post Anthropic’s Claude will soon help you make sense of your Apple Watch health data appeared first on Digital Trends.

Source:: Digital Trends

By Deepti Pathak OpenAI has announced ChatGPT Health, a new feature inside ChatGPT that focuses only on health and…
The post What Is ChatGPT Health & How To Join the Waitlist? appeared first on Fossbytes.

Source:: Fossbytes

By Moinak Pal A team of researchers in China has just pulled the curtain back on a new sodium-sulfur battery design that could fundamentally change the math on energy storage. By leaning into the very chemistry that has historically made sulfur a headache for engineers, they have managed to build a cell that is incredibly cheap to make […] The post Chinese researchers develop high-voltage sodium–sulfur battery that could challenge lithium batteries appeared first on Digital Trends.

Source:: Digital Trends