Chinas open-weight AI models surge past US rivals in usage as Silicon Valley scrambles to respond

Chinas open-weight AI models surge past US rivals in usage as Silicon Valley scrambles to respond
30 articles·19 sources·updated about 2 hours ago·View in graph

Story Timeline

24 hours · 2 summary articles

  1. Chinas open-weight AI models surge past US rivals in usage as Silicon Valley scrambles to respond

  2. China's Kimi-K3 AI model sparks global tech selloff at Shanghai conference: Xi urges UN-led governance

    Continuation
science & techestoniachinaunited states of america
AI-generated · Hosted in Europe

China's latest AI breakthrough is challenging Silicon Valley's dominance in the field. Chinese labs like Moonshot AI are leading the market for cheap, customizable intelligence, threatening to turn America's prestige models into expensive niche products.

Moonshot AI's Kimi K3 model, released this week, has shown performance levels close to those of leading U.S. models like Anthropic's Fable and OpenAI's GPT-5.6. This development is causing concern in Silicon Valley, as businesses can use cheaper systems for routine tasks, reserving premium models for more complex problems.

According to a report by Axios, Chinese models now occupy the top five spots by weekly token usage on OpenRouter, a major marketplace for AI systems. These models are "open-weight," allowing users to download, customize, and run them on their own systems. This shift is making it easier for businesses to adopt cheaper alternatives.

Industry experts are taking note of this shift. Augusto Marietti, CEO of Kong, told Axios that open-weight use has surged over the past quarter as flagship models are "too expensive." Raffi Krikorian, CTO of Mozilla, compared using frontier AI for everyday work to "driving a Ferrari to Whole Foods."

The rapid advancement of Chinese AI models is also raising concerns about cybersecurity. The British AI Security Institute (AISI) reported that open models now trail closed systems in cyber capabilities by only four to seven months, down from six to ten months earlier this year. This is causing alarm among cyber defenders, who have less time to prepare for new types of attacks.

In a significant move, Alibaba's chip design unit T-Head announced at the World AI Conference in Shanghai that it is open-sourcing SAIL, the full software stack for its Zhenwu series of AI chips. This move is designed to lower migration barriers for developers currently locked into Nvidia's CUDA ecosystem.

Chinese President Xi Jinping has been pushing for a more secure AI strategy and has called for stronger international coordination through the United Nations. At the World AI Conference in Shanghai, Xi announced the launch of the "World Artificial Intelligence Cooperation Organization" and plans to provide 5,000 AI training slots for Global South countries.

In response to the growing threat from Chinese AI models, some American companies are scrambling to fight back. Thinking Machines, a startup launched by former OpenAI CTO Mira Murati, made its debut this week with an open-weight model built for deep customization. Nvidia is also expanding its Nemotron family of open models, betting that customizable AI will drive more demand for the company's chips and software.

SpaceX, led by Elon Musk, is also making moves in the AI space. The company is negotiating with the Pentagon to provide AI data centers worth several billion dollars. SpaceX aims to sell computing power to customers at lower prices.

The shift towards cheaper, customizable AI models is not just a technological race but also has significant economic and geopolitical implications. As China continues to advance its AI capabilities, the global landscape of AI governance and market dynamics is likely to undergo significant changes.

Share

Follow us for live European news

Source Intelligence
19 sources10 countries
Geographic Origin12 located
  • 3
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1

7 further sources not geolocated

Political Spectrum7 mapped

Articles

Live From Europe

Nações Unidas da Inteligência Artificial: Xi Jinping defende supervisão global para reforçar segurança O líder chinês quer uma abordagem centrada nas pessoas, defende o princípio da abertura e do benefício mútuo e uma abordagem baseada no verdadeiro multilateralismo.

publico · about 3 hours ago

Live From Europe

India achieves milestone with launch of first private-sector orbital rocket PM Narendra Modi says successful launch will encourage countless youngsters to dream bigger and innovate fearlessly.

aljazeera · about 3 hours ago

✨ This year marks 25 years since the first international meeting of investigative journalists, GIJC. Join us starting October 26th to celebrate this anniversary with #IJWeek to network, fundraise, and learn. Early bird tickets on sale now!

✨ This year marks 25 years since the first international meeting of investigative journalists, GIJC. Join us starting October 26th to celebrate this anniversary with #IJWeek to network, fundraise, and learn. Early bird tickets on sale now!

bluesky_GIJN · about 3 hours ago

Shanghai Xingshu launches first constellation of its space-based computing project reut.rs/4b2kv9L

Shanghai Xingshu launches first constellation of its space-based computing project reut.rs/4b2kv9L

bluesky_Reuters · about 3 hours ago

Live From Europe

Chinas new World Artificial Intelligence Cooperation Organization is President Xis clearest play yet for a parallel AI order At the World AI Conference in Shanghai, Xi Jinping announced 5,000 AI training slots for Global South countries and the launch of the "World Artificial Intelligence Cooperation Organization." Cooperation centers with ASEAN, the African Union, BRICS, and other alliances are planned to follow. China is systematically building a parallel AI governance structure outside Western influence. The article Chinas new World Artificial Intelligence Cooperation Organization is President Xis clearest play yet for a parallel AI order appeared first on The Decoder.

the decoder · about 3 hours ago

Live From Europe

Alibaba open-sources its AI chip software stack at WAIC, targeting Nvidias CUDA lock-in Alibabas chip design unit T-Head announced at the World AI Conference in Shanghai on Saturday that it is open-sourcing SAIL, the full software stack for its Zhenwu series of AI chips. The move is designed to lower migration barriers for developers currently locked into Nvidias CUDA ecosystem. T-Head said programmers can adapt SAIL to mainstream […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

How Googles New Gemini Rates Work and How to Track Your Usage Now that Google has changed how its usage quotas are tallied, you might not get as many AI responses as you did before.

wired.com · about 3 hours ago

Live From Europe

SpaceX a picco del 45 in un mese dopo la quotazione da record. E per Musk arriva il test delle vendite dei grandi soci A poco più di un mese dallofferta pubblica iniziale più grande della storia, la luna di miele tra Wall Street e SpaceX sembra già finita. Venerdì il titolo della società di Elon Musk ha chiuso a 123,99 dollari: sotto i 135 dollari del prezzo di collocamento, un debutto da record che aveva valutato lazienda 1.750 […] Larticolo SpaceX a picco del 45 in un mese dopo la quotazione da record. E per Musk arriva il test delle vendite dei grandi soci proviene da Il Fatto Quotidiano.

il fatto quotidiano · about 3 hours ago

Live From Europe

Traders are increasingly betting against SpaceX just weeks after IPO Shares of Elon Musks rockets and AI company have slipped below their listing price for the first time

financial times · about 3 hours ago

Live From Europe

The Most Important Words in the Battle Over AI In the battle over AI regulation — from Washington, D.C. to Silicon Valley — one term has become unexpectedly controversial: AI safety. After all, who doesnt believe in making these tools safe? But depending on where people stand on the issue, AI safety has come to signal wildly different things. Some are concerned about existential […]

politico.eu · about 3 hours ago

The benchs AI balancing act Judges are becoming AI firewalls, setting guidelines for lawyers and litigants entering their courtrooms and reprimanding those who fail to vet their research or cite fake AI-generated cases.But judges are also navigating how to use the tech themselves.Why it matters: As attorney sanctions mount for irresponsible AI use, judges chambers are at the center of the push and pull between innovation and caution. Some early adopters see AI use as a catalyst for justice — while other jurists remain staunchly skeptical."AI has the possibility to be an incredible value adder and force maximizer in the federal judiciary," says Amy Cyphert, a West Virginia University College of Law professor. "But it also has the potential to undermine public confidence in the judiciary at a time where its already at a historic low." By the numbers: 60 of 112 judges surveyed by Northwestern researchers reported using at least one AI tool. Just over 22 used AI tools weekly or daily.Cypherts own research found judges worry about not only the risks of AI hallucinations but also the privacy and confidentiality implications.Yes, but: AI adoption is outpacing training, the judges reported in the survey. To fill those gaps, Magistrate Judge Maritza Braswell of the U.S. District Court of Colorado founded the Judicial AI Consortium (JAIC), alongside Judge Scott Schlegel of Louisianas circuit court and U.S. District Judge Xavier Rodriguez. "Because the technology is moving very fast, I think the tendency tends to be to start from a place of fear," she says. So, "weve really created the space where we start from a place of knowledge."Friction point: U.S. District Judge Liam OGrady of the Eastern District of Virginia tells Axios that before he retired in 2023, judges were fascinated by AI — but they also "had pretty grave doubts about its reliability."Lawyers, he says, had begun turning to the tools — and "it was clear that mistakes were being made right away."Case in point: Last month, a federal judge in Mississippi punished four lawyers and canceled a civil trial after both sides cited fake, AI-generated cases. And its not just lawyers who have faced heat.Damien Charlotin, a senior research fellow at HEC Paris, has identified more than 1,700 cases where generative AI hallucinated. On Thursday, a federal judge in Michigan accused the government of likely citing an AI-generated case in an immigration-related filing. While Chief Judge Hala Jarbou, a Trump appointee, did not impose sanctions, she wrote that "the Government must ensure its future filings with this Court do not include nonexistent case law."The solution, JAICs judicial co-founders who spoke with Axios emphasized, is education."We didnt shut down Zoom court because one lawyer showed up as a cat," Schlegel says. Rodriguez adds that its "a work in progress by everybody — lawyers and judges alike."Plus, Braswell worries an outsized focus on hallucinations could distract from other ethical considerations, like bias and access to justice.Threat level: Schlegel adds that judges approach lawyers or pro se litigants use of AI differently than how they think about using it in their chambers."We need to be much more thoughtful in the way that we use these tools," he says, adding, "When we sign an order, if theres something wrong in it, thats the law."The bottom line: Judges are learning a fast-moving technology from a seat where getting it wrong is most costly."I started practice in the 70s with the typewriter," OGrady says, adding, "Theres been such a dramatic jump in technology."He says its "like learning a new language when youre 60 or 70 years old — Its not easy. Weve got a whole generation of judges who this is just completely new to, and theyre cautious and skeptical."Go deeper: AI tells lawyers how judges are likely to rule

The benchs AI balancing act Judges are becoming AI firewalls, setting guidelines for lawyers and litigants entering their courtrooms and reprimanding those who fail to vet their research or cite fake AI-generated cases.But judges are also navigating how to use the tech themselves.Why it matters: As attorney sanctions mount for irresponsible AI use, judges chambers are at the center of the push and pull between innovation and caution. Some early adopters see AI use as a catalyst for justice — while other jurists remain staunchly skeptical."AI has the possibility to be an incredible value adder and force maximizer in the federal judiciary," says Amy Cyphert, a West Virginia University College of Law professor. "But it also has the potential to undermine public confidence in the judiciary at a time where its already at a historic low." By the numbers: 60 of 112 judges surveyed by Northwestern researchers reported using at least one AI tool. Just over 22 used AI tools weekly or daily.Cypherts own research found judges worry about not only the risks of AI hallucinations but also the privacy and confidentiality implications.Yes, but: AI adoption is outpacing training, the judges reported in the survey. To fill those gaps, Magistrate Judge Maritza Braswell of the U.S. District Court of Colorado founded the Judicial AI Consortium (JAIC), alongside Judge Scott Schlegel of Louisianas circuit court and U.S. District Judge Xavier Rodriguez. "Because the technology is moving very fast, I think the tendency tends to be to start from a place of fear," she says. So, "weve really created the space where we start from a place of knowledge."Friction point: U.S. District Judge Liam OGrady of the Eastern District of Virginia tells Axios that before he retired in 2023, judges were fascinated by AI — but they also "had pretty grave doubts about its reliability."Lawyers, he says, had begun turning to the tools — and "it was clear that mistakes were being made right away."Case in point: Last month, a federal judge in Mississippi punished four lawyers and canceled a civil trial after both sides cited fake, AI-generated cases. And its not just lawyers who have faced heat.Damien Charlotin, a senior research fellow at HEC Paris, has identified more than 1,700 cases where generative AI hallucinated. On Thursday, a federal judge in Michigan accused the government of likely citing an AI-generated case in an immigration-related filing. While Chief Judge Hala Jarbou, a Trump appointee, did not impose sanctions, she wrote that "the Government must ensure its future filings with this Court do not include nonexistent case law."The solution, JAICs judicial co-founders who spoke with Axios emphasized, is education."We didnt shut down Zoom court because one lawyer showed up as a cat," Schlegel says. Rodriguez adds that its "a work in progress by everybody — lawyers and judges alike."Plus, Braswell worries an outsized focus on hallucinations could distract from other ethical considerations, like bias and access to justice.Threat level: Schlegel adds that judges approach lawyers or pro se litigants use of AI differently than how they think about using it in their chambers."We need to be much more thoughtful in the way that we use these tools," he says, adding, "When we sign an order, if theres something wrong in it, thats the law."The bottom line: Judges are learning a fast-moving technology from a seat where getting it wrong is most costly."I started practice in the 70s with the typewriter," OGrady says, adding, "Theres been such a dramatic jump in technology."He says its "like learning a new language when youre 60 or 70 years old — Its not easy. Weve got a whole generation of judges who this is just completely new to, and theyre cautious and skeptical."Go deeper: AI tells lawyers how judges are likely to rule

axios · about 3 hours ago

Live From Europe

Women-led start-ups attract just 7 of Swiss venture capital funding Start-ups led by women continue to have a limited presence in the Swiss venture capital market. In the first half of 2026, just over 7 of venture capital was channelled to companies with female CEOs. The proportion of funding allocated to women-led teams has therefore remained at a low level for years, according to a new study by the investors association SECA. Given that gender equality has long been a topic of discussion and there are now a number of funding initiatives in place, this is astonishing, said Stefan Kyora, co-author of the study, at a media event. It is also surprising that, in the start-up world, most management teams remain homogeneous – and thus predominantly male. Specifically, according to the study, companies with all-male management teams received more than 88 of the total capital invested in the first half of the year. + How we produce English news Translated from French, reviewed by an English Department journalist.

swissinfo · about 3 hours ago

Live From Europe

Face AI upgrades its video face swap tool with faster processing and better tracking Face AI, a Los Angeles-based face swap platform, announced a major update to its video face swap feature on Friday. The update improves facial tracking, expression preservation, and scene stability across changing lighting, camera angles, and partial face occlusions like glasses and hats. Queued videos now process in under a minute. The tool targets social […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

Public sale of Volta SKAI B-tower begins: the most expensive apartment purchase in the history of the Estonian real estate market was concluded during the pre-sale - The Baltic Times Real estate developer Endover is launching the public sale of apartments and commercial premises in the second tower of the Volta SKAI h......

baltic times · about 3 hours ago

Live From Europe

NORDSPACE secures EUR 1.2 million financing from Signet Bank for the development of smart storage facilities in Riga - The Baltic Times SIA Nordspace-Jurkalnes has secured EUR 1.2 million in financing from Signet Bank to implement the second phase of the smart storage com......

baltic times · about 3 hours ago

Live From Europe

Gunārs Ķirsons launches new DABA restaurant concept with EUR 4.5 mllion investment - The Baltic Times RIGA – Latvian entrepreneur and LIDO founder Gunārs Ķirsons is launching DABA, a new fast-casual restaurant concept, with the firs......

baltic times · about 3 hours ago

Live From Europe

Boeing: Global Commercial Fleet Will Top 50,000 Airplanes in 20 Years as Passenger Traffic Doubles

gdeltproject.org · about 3 hours ago

Live From Europe

Black Lake Technologies Shortlisted as SAIL Award TOP30 Finalist and Selected as Global Industrial AI Flagship Case, Showcasing Latest Industrial Agent at WAIC 2026

gdeltproject.org · about 3 hours ago

🇮🇳 🚀 Indias first privately built orbital rocket took its maiden flight on Saturday, its manufacturer said, marking a significant step for the South Asian giant as it eyes a bigger slice of the global space economy ➡️ u.afp.com/Sr5M

🇮🇳 🚀 Indias first privately built orbital rocket took its maiden flight on Saturday, its manufacturer said, marking a significant step for the South Asian giant as it eyes a bigger slice of the global space economy ➡️ u.afp.com/Sr5M

bluesky_AFP English · about 3 hours ago

Live From Europe

Open-weight models now match frontier cyber performance from just four months ago at a fraction of the cost The British AI Security Institute warns that open-weight models like GLM-5.2 and DeepSeek V4-Pro now trail closed frontier models in cyber capabilities by four to seven months. At the start of 2025, the gap was still six to ten months. It also found that safety measures on open models are largely ineffective, leaving defenders less time to prepare. The article Open-weight models now match frontier cyber performance from just four months ago at a fraction of the cost appeared first on The Decoder.

the decoder · about 3 hours ago

Live From Europe

A French startup built a radiology viewer from scratch with AI at its core. Moffitt Cancer Center is already using it. Raidium, a Paris and Silicon Valley-based radiology startup, has launched its AI-native imaging platform in the US at Moffitt Cancer Center, one of the countrys leading oncology research institutions. The platform, called Raidium Read, replaced Moffitts legacy radiomics applications and is currently available for clinical trials and research use. FDA 510(k) clearance is expected before […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

Meta patented an AI system that listens to your voice all day and tracks your mood Meta has been granted a patent for an AI system that continuously records a users voice, transcribes it, and feeds the audio through a machine learning model to detect their emotional state. The patent, published on July 2, describes a device that listens for audible communications including sighs, laughter, and vocal tone, and combines them […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

Nebius raised $775 million by borrowing against its GPUs. It has $40 billion more contracts to securitise. Nebius raised $775 million in its first secured debt facility, borrowing against deployed GPU infrastructure and contracted cash flows from an investment-grade customer. The facility matures on October 31, 2030, and is priced at SOFR + 2.50, roughly 6.8 at current rates. Together with the customer agreements cash flows, the facility covers more than 100 […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

The White House is now deciding who gets access to frontier AI models, not the labs The Trump administration is now dictating which companies and entities get access to frontier AI models from Anthropic and OpenAI, CNBC reported on Friday, citing two people familiar with the matter. Until now, the labs made that decision themselves. Anthropic controlled access to its Mythos cybersecurity model through an initiative called Project Glasswing. OpenAI ran […] This story continues at The Next Web

the next web · about 3 hours ago

Live From Europe

Xi pushes secure AI strategy as China expands global initiative Chinese President Xi Jinping unveiled a new international AI cooperation platform while urging countries to develop artificial intelligence under stricter security and governance standards. Speaking at the World AI Conference in Shanghai, Xi argued that AI should remain under human oversight and called for stronger international coordination through the United Nations.

yenisafak · about 3 hours ago

Bienvenue dans le Splinternet, lère de la fragmentation du net Lutopie dInternet comme réseau ouvert et universel a fait long feu. Plusieurs pays ont mis au point des systèmes qui leur permettent de se couper du réseau mondial pour fonctionner en circuit (presque) fermé, des intranets nationaux.

Bienvenue dans le Splinternet, lère de la fragmentation du net Lutopie dInternet comme réseau ouvert et universel a fait long feu. Plusieurs pays ont mis au point des systèmes qui leur permettent de se couper du réseau mondial pour fonctionner en circuit (presque) fermé, des intranets nationaux.

courrier international · about 3 hours ago

Live From Europe

China will mit günstiger KI die Dominanz der USA brechen [premium] Staatschef Xi Jinping will mit offenen KI-Modellen die USA bei Künstlicher Intelligenz überholen. Dahinter steckt nicht nur technologisches, sondern auch wirtschaftliches Kalkül.

die presse · about 3 hours ago

Live From Europe

Chips are nearly five times as volatile as the broader market Semiconductor stock roulette

financial times · about 3 hours ago

Live From Europe

Global stocks slide amid Iran-US tensions and semiconductor sell-off Global markets fell as escalating US-Iran tensions and heavy selling in semiconductor stocks weighed on investor sentiment. Chipmakers remained under pressure over AI investment returns, while rising oil prices raised inflation concerns despite resilient US economic data.

yenisafak · about 3 hours ago

Befektetésekről – alfától az omegáig Ismét új prémium hírlevelet mutatunk be a hvg360 előfizetőinek, ezúttal az OTP Global Markets szakértőitől.

Befektetésekről – alfától az omegáig Ismét új prémium hírlevelet mutatunk be a hvg360 előfizetőinek, ezúttal az OTP Global Markets szakértőitől.

hvg.hu · about 3 hours ago