China’s DeepSeek launches next-gen AI model. Here’s what makes it different
Anna Barclay | Getty Images News | Getty Images
Chinese startup DeepSeek’s latest experimental model promises to increase efficiency and improve AI’s ability to handle a lot of information at a fraction of the cost, but questions remain over how effective and safe the architecture is.
DeepSeek sent Silicon Valley into a frenzy when it launched its first model R1 out of nowhere last year, showing that it’s possible to train large language models (LLMs) quickly, on less powerful chips, using fewer resources.
The company released DeepSeek-V3.2-Exp on Monday, an experimental version of its current model DeepSeek-V3.1-Terminus, which builds further on its mission to increase efficiency in AI systems, according to a post on the AI forum Hugging Face.
“DeepSeek V3.2 continues the focus on efficiency, cost reduction, and open-source sharing,” Adina Yakefu, Chinese community lead at Hugging Face, told CNBC. “The big improvement is a new feature called DSA (DeepSeek Sparse Attention), which makes the AI better at handling long documents and conversations. It also cuts the cost of running the AI in half compared to the previous version.”
“It’s significant because it should make the model faster and more cost-effective to use without a noticeable drop in performance,” said Nick Patience, vice president and practice lead for AI at The Futurum Group. “This makes powerful AI more accessible to developers, researchers, and smaller companies, potentially leading to a wave of new and innovative applications.”
The pros and cons of sparse attention
An AI model makes decisions based on its training data and new information, such as a prompt. Say an airline wants to find the best route from A to B, while there are many options, not all are feasible. By filtering out the less viable routes, you dramatically reduce the amount of time, fuel and, ultimately, money, needed to make the journey. That is exactly sparse attention does, it only factors in data that it thinks is important given the task at hand, as opposed to other models thus far which have crunched all data in the model.
“So basically, you cut out things that you think are not important,” said Ekaterina Almasque, the cofounder and managing partner of new venture capital fund BlankPage Capital.
Sparse attention is a boon for efficiency and the ability to scale AI given fewer resources are needed, but one concern is that it could lead to a drop in how reliable models are due to the lack of oversight in how and why it discounts information.
“The reality is, they [sparse attention models] have lost a lot of nuances,” said Almasque, who was an early supporter of Dataiku and Darktrace, and an investor in Graphcore. “And then the real question is, did they have the right mechanism to exclude not important data, or is there a mechanism excluding really important data, and then the outcome will be much less relevant?”
This could be particularly problematic for AI safety and inclusivity, the investor noted, adding that it may not be “the optimal one or the safest” AI model to use compared with competitors or traditional architectures.
DeepSeek, however, says the experimental model works on par with its V3.1-Terminus. Despite speculation of a bubble forming, AI remains at the centre of geopolitical competition with the U.S. and China vying for the winning spot. Yakefu noted that DeepSeek’s models work “right out of the box” with Chinese-made AI chips, such as Ascend and Cambricon, meaning they can run locally on domestic hardware without any extra setup.
DeepSeek also shared the actual programming code and tools needed to use the experimental model, she said. “This means other people can learn from it and build their own improvements.”
But for Almasque, the very nature of this means the tech may not be defensible. “The approach is not super new,” she said, noting the industry has been “talking about sparse models since 2015” and that DeepSeek is not able to patent its technology due to being open source. DeepSeek’s competitive edge, therefore, must lie in how it decides what information to include, she added.
The company itself acknowledges V3.2-Exp is an “intermediate step toward our next-generation architecture,” per the Hugging Face post.
As Patience pointed out, “this is DeepSeek’s value prop all over: efficiency is becoming as important as raw power.”
“DeepSeek is playing the long game to keep the community invested in their progress,” Yakefu added. “People will always go for what is cheap, reliable, and effective.”
The Walt Disney Co. signage on the floor of the New York Stock Exchange (NYSE) in New York, US, on Wednesday, May 7, 2025.
Michael Nagle | Bloomberg | Getty Images
The Walt Disney Company sent a cease and desist letter to Character.AI last week, warning the artificial intelligence startup to stop using copyrighted characters without authorization, a Disney spokesperson confirmed to CNBC on Tuesday.
A spokesperson for Character.AI said it removed the characters mentioned in the letter, and that “it’s always up to rightsholders to decide how people may interact with their IP.”
The spokesperson acknowledged that while some characters on its platform are completely original creations, others are “inspired by existing characters that people love.”
“We want to partner with the industry and rightsholders to empower them to bring their characters to our platform,” the Character.AI spokesperson told CNBC. “Our goal is to give IP owners the tools to create controlled, engaging and revenue-generating experiences from deep fandom for their characters and stories, expanding their reach using our new, interactive format.”
The letter serves as the latest example of how media companies like Disney are working to protect their intellectual property during the AI boom.
Disney is already involved in an ongoing lawsuit against AI image creator Midjourney, alleging that the company improperly used and distributed AI-generated characters from movies like “Cars,” “Toy Story,” “Shrek,” “The Avengers” and others.
Axios first reported the cease and desist letter.
Character.AI allows users to create and interact with character-based chatbots. Google inked a $2.7 billion licensing deal with Character.AI and hired its founders in 2024, and the startup became embroiled in a wrongful death lawsuit that same year.
The family of Sewell Setzer III, a 14-year-old boy in Florida, alleged he committed suicide after he became addicted to talking with a number of AI chatbots on the app. One of the chatbots was named Daenerys Targaryen, or Dany, who is a character in the show “Game of Thrones,” according to the lawsuit.
Character.AI is not the only AI company that’s faced scrutiny over its approach to IP.
Earlier this month, a federal judge preliminarily approved Anthropic’s offer to pay $1.5 billion to settle a class action lawsuit with a group of authors, who claimed that the company had illegally downloaded their books and others from pirated databases.
If you are having suicidal thoughts or are in distress, contact the Suicide & Crisis Lifeline at 988 for support and assistance from a trained counselor.
WATCH: Disney likely to spin off ESPN and ABC post-Iger, says LightShed’s Rich Greenfield
Daniel Rausch, vice president of Alexa and Echo, announces the Echo Studio and Echo Dot Max during an Amazon event showcasing new products in New York City, U.S., September 30, 2025.
Kylie Cooper | Reuters
Amazon on Tuesday unveiled four new smart speakers and voice-activated displays that are revamped with Alexa+, its personal assistant that’s powered by generative artificial intelligence.
The company debuted the Echo Dot Max, a revamped version of its compact smart speaker, which costs $99.99. Amazon also unveiled a new Echo Show 8 and Echo Show 11, priced at $179.99 and $219.99, respectively.
There’s also a new version of the Echo Studio, a larger, higher-end model with a more powerful speaker, priced at $219.99.
All the devices are available for preorder on Tuesday, and users will get Alexa+ early access “out of the box,” Amazon said. The Echo Dot Max and Echo Studio ship Oct. 29, while the Echo Show 8 and Echo Show 11 ship Nov. 12.
The devices were launched at Amazon’s fall hardware bonanza, held in New York. They’re the first batch of revamped products under the leadership of Panos Panay, a former Microsoft hardware leader who joined Amazon in 2023.
It’s also the first set of Amazon hardware to integrate the company’s long-awaited Alexa+, which debuted in February and has slowly rolled out in early access for some users.
“These are the most powerful Echo devices we have ever created,” Panay said on stage at the event. “Custom silicon, advanced sensors, our best microphones and sound, noise cancellation, understanding the user, faster than anything we’ve ever delivered before. They’re also beautifully designed to fade into the background.”
Alongside a revamped look, Amazon added new AZ3 and AZ3 Pro chips for edge processing to the devices, which are faster, more powerful and have “AI built right in,” said Daniel Rausch, the head of Amazon’s Alexa and Echo businesses.
Panos Panay, head of Amazon’s Devices and Services team, introduces Echo during an Amazon product event in the Manhattan borough of New York City on September 30, 2025. Amazon announced its next generation of Kindle, Ring, Blink, Fire TV, and Echo devices.
Charly Triballeau | Afp | Getty Images
The devices also feature a so-called Omnisense platform that gives Alexa “better contextual awareness,” Rausch said. It allows the Echo Show to be able to recognize users and serve up personalized insights, like an analysis of how they slept last night or alert users if they left their front door unlocked after midnight.
Amazon faces growing pressure to update its hardware and software for the generative AI age following the success of rivals such as OpenAI’s ChatGPT and Google’s Gemini. Meta also has its Ray-Ban Meta glasses, which use its Llama large language model to answer spoken questions from the user.
Amazon is also looking beyond Alexa or Echo smart speakers for opportunities in device growth.
The company in July confirmed it’s acquiring AI wearables startup Bee, which makes a wristband that can record and transcribe conversations.
Nvidia CEO Jensen Huang attends the “Winning the AI Race” Summit in Washington D.C., U.S., July 23, 2025.
Kent Nishimura | Reuters
Nvidia shares reached a fresh record on Tuesday, climbing almost 3% and lifting the chipmaker’s market cap past $4.5 trillion.
The stock is now up about 39% for the year, and continues to attract investors as Nvidia steps up its pace of deal-making, cementing its position at the center of the artificial intelligence boom.
OpenAI said last week that Nvidia would take an equity stake worth up to $100 billion in the AI startup, and would build hundreds of billions of dollars worth of data centers filled with Nvidia graphics processing units. OpenAI then announced five massive new data centers with Oracle that are expected to be filled with hundreds of thousands of GPUs. The whole “Stargate” project will cost $500 billion, the companies said.
Nvidia CEO Jensen Huang says Nvidia’s products comprise about 70% of the spending on a new AI data center.
Analysts at Citi on Tuesday raised their price target on Nvidia from $200 to $210, citing an increased forecast for AI infrastructure spending after the OpenAI announcements.
“We believe OpenAI came to Nvidia asking for help as Nvidia has a very compelling product, and as the number of users and compute being consumed per user basis is growing,” Citi analyst Atif Malik wrote in the note.
OpenAI is far from alone, as Meta, Google and others are also dramatically ramping up their infrastructure spending.
CoreWeave, a cloud provider that includes Nvidia as a large shareholder, said Tuesday it had reached a deal to supply Meta with $14.2 billion in AI infrastructure services.
Nvidia’s stock is outperforming all of its megacap peers so far this year except for chipmaker Broadcom, which is up about 40%, similarly boosted by OpenAI.
WATCH: Worried about an AI bubble? Trader says she prefers hard assets, especially silver
-
Sports3 years ago‘Storybook stuff’: Inside the night Bryce Harper sent the Phillies to the World Series
-
Sports1 year agoStory injured on diving stop, exits Red Sox game
-
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
-
Environment12 months agoHere are the best electric bikes you can buy at every price level in October 2024