Chinese DoorDash rival smashes open source record with 1.6-trillion-parameter LLM with a 1-million-context-token model crafted without Nvidia hardware
Meituan has unveiled LongCat-2.0, a groundbreaking 1.6-trillion-parameter language model developed entirely on domestic chips, marking a significant achievement in the open-source AI landscape while completely bypassing Nvidia hardware.
WPN Brief
- What Happened
Meituan has unveiled LongCat-2.0, a groundbreaking 1.6-trillion-parameter language model developed entirely on domestic chips, marking a significant achievement in the open-source AI landscape while completely bypassing Nvidia hardware.
- Why It Matters
This development positions Meituan as a formidable player in the AI sector, showcasing its capabilities in creating advanced models without reliance on established hardware suppliers, which could enhance its competitive edge and innovation potential.
- The Bigger Picture
The release of LongCat-2.0 reflects a broader trend in the tech industry where companies are increasingly exploring alternatives to Nvidia's dominance, as seen in various initiatives aimed at developing independent AI technologies and addressing the growing demand for advanced AI solutions across multiple sectors.
Related Reports
More coverage on this story
10 reports across the wire
Tencent's Apache-licensed Hy3 takes on GLM-5.2 at half the size — and wins everywhere except coding
Tencent has launched the Hy3 model, a 295-billion-parameter Mixture-of-Experts (MoE) model, under the Apache 2.0 license, addressing previous licensing restrictions that limited access for enterprises in the EU, UK, and South Korea. This release positions Hy3 as a competitive alternative to GLM-5.2, particularly in various applications except coding.
Huawei just dropped a major AI chip surprise — but Nvidia should really be paying attention to the other big news from its rival
Huawei has announced the Ascend 950DT AI chip, highlighting its rapid development in artificial intelligence technology and its growing ecosystem. This announcement is seen as a significant move in the competitive landscape against Nvidia.
Nvidia Gets Into the PC Market With New Chip | Bloomberg Tech 6/1/2026
Nvidia has officially entered the PC market with the announcement of its new RTX Spark Superchip, which will be integrated into laptops and desktop computers from major brands like Dell and Lenovo this fall, as revealed by CEO Jensen Huang during the Computex trade show.
Microsoft launches MXC, an OS-level sandbox for AI agents, with OpenAI and Nvidia already on board
Microsoft has launched Microsoft Execution Containers (MXC) at its Build 2026 conference, introducing an OS-level sandbox designed to enhance the security and management of AI agents by allowing developers to set access boundaries enforced by the operating system. This initiative aims to address concerns about AI agents malfunctioning and causing security risks.
The Startups Building on Nvidia Compute
Nvidia is positioning itself as a key player in the AI sector, moving beyond traditional chip sales to support a burgeoning ecosystem of startups developing AI technologies, including agents, humanoid robots, and robotaxis. CEO Jensen Huang's recent message to investors emphasizes this shift, highlighting the role of venture capitalists like Sarah Guo of Conviction in fostering innovation.
Japan reveals new Noetra plan to flood the country with 10 million robots by 2040 — including work in the nursing, food and drink sectors
Japan has announced an ambitious plan to deploy 10 million robots by 2040, leveraging Noetra AI models to tackle labor shortages in critical sectors such as nursing and food production. This initiative aims to enhance productivity and address demographic challenges faced by the country.
On-device AI agents hit a hard memory limit. Apple's new architecture routes around it.
Apple has announced a new architecture for on-device AI models that circumvents the memory limitations of DRAM, allowing for larger parameter counts. This development was highlighted during the Worldwide Developers Conference (WWDC) 2026, where Apple introduced its AFM 3 family of models, including a 20-billion-parameter version that utilizes NAND flash storage instead of DRAM.
Chinese firm's fiber trial hits staggering 51.3 Tb/s over 128 miles without signal regeneration — could this be an end to bandwidth bottlenecks?
YOFC has successfully demonstrated a groundbreaking fiber optic transmission capability, achieving 51.3 Tb/s over a distance of 206.5 km without the need for signal regeneration. This advancement is poised to significantly alleviate the increasing demands on global AI networking infrastructure.
Nvidia Is Bigger, Stronger, and Trying to Slay the Laptop Dragon Again
Nvidia has made a significant move by launching its new RTX Spark superchip, marking its re-entry into the laptop market and aiming to compete with established players like Intel and AMD. This development is part of Nvidia's broader strategy to enhance its product offerings and capitalize on the growing demand for AI-capable computing solutions.
Trump Discussed Nvidia Chips With Xi Jinping | Bloomberg Tech 5/15/2026
During a recent summit in Beijing, former President Donald Trump discussed Nvidia's H200 chips and artificial intelligence guardrails with Chinese President Xi Jinping. This meeting highlights ongoing diplomatic efforts between the U.S. and China to address critical technology and trade relations.