Senior AI Journalist
Alibaba Unveils Qwen 3.8, Challenging Frontier AI Models
Alibaba has officially unveiled Qwen 3.8, its latest open-weight AI model boasting an impressive 2.4 trillion parameters. The Qwen team claims this new iteration matches the performance of many frontier models currently on the market, positioning it just behind Anthropic’s Fable 5. This release signals Alibaba’s significant commitment to advancing open-source AI capabilities and democratizing access to powerful large language models.
A preview of Qwen 3.8 is now accessible through Alibaba’s Token Plan, Qoder, and QoderWork platforms, offered at a competitive 10 percent of the standard price. While the full open weights are anticipated to be released “soon,” this initial access allows developers and researchers to experiment with the model’s capabilities and provide valuable feedback. The strategic pricing suggests an aggressive push to gain market share and promote wider adoption of Alibaba’s AI ecosystem.
Qwen 3.8 is expected to significantly outperform its predecessor, Qwen 3.7-Max, particularly in complex domains like coding and high-level productivity tasks such as full-stack development, data analysis, and office workflows. Developer Shuai Bai highlighted that this is also the team’s first multimodal model with over a trillion parameters, capable of processing images, videos, and documents alike. While specific benchmark results are yet to be released, the early indicators point to a versatile and robust AI model designed for a broad range of applications.
Source: The Decoder
Meta Explores $10 Billion Compute Partnership with Anthropic
Meta is reportedly engaged in high-level discussions with Anthropic to lease a substantial portion of its data center compute capacity. This potential deal, valued at up to $10 billion over a two-year period, could mark a significant strategic partnership between two AI giants. The negotiations, first reported by the New York Times, indicate Meta’s interest in monetizing its vast AI infrastructure and Anthropic’s urgent need for additional computational resources to meet surging demand for its Claude models.
Anthropic, a leading AI research company, initially proposed the arrangement to Meta in June. While Meta is still reviewing the intricate details of the proposal, a successful agreement would provide a new revenue stream for the social media conglomerate. CEO Mark Zuckerberg has previously hinted at the possibility of selling excess compute capacity if the company’s internal AI demands did not fully utilize its expansive data centers, and this deal could validate that strategy on an unprecedented scale.
Meta’s ambitious investment in AI infrastructure is well-documented, with plans to spend up to $145 billion this year across various AI initiatives. This includes developing new revenue sources such as paid AI subscriptions like “Meta One” and even exploring futuristic endeavors like space-based solar power for its data centers. For Anthropic, the influx of compute power is critical, especially as demand for its Claude Code has surged, following its recent $45 billion deal with SpaceXAI. Despite this, Anthropic maintains a long-term vision of building its own data centers, a goal it is actively pursuing by recruiting former Google executives.
Source: The Decoder
Jensen Huang’s Japan Visit Fuels Nation’s Physical AI Ambitions
NVIDIA CEO Jensen Huang recently concluded a impactful two-day visit to Tokyo, where he embarked on a mission to fortify partnerships within Japan’s industrial and chip-supply sectors. Fresh from keynotes in Taiwan and South Korea, Huang’s agenda in Japan was comprehensive, yielding agreements spanning the nation’s entire tech ecosystem. His clear message: NVIDIA is keenly focused on the factory floor, robotics, and intelligent machines, urging Japan to lead the charge in defining AI’s next frontier in the physical world.
Decades after a pivotal $5 million investment from Sega saved a struggling NVIDIA, the relationship between NVIDIA and Japan’s industrial titans has re-emerged with renewed significance. This time, the collaboration is centered on ushering in the era of physical AI. Key initiatives unveiled include Noetra—Japan’s sovereign AI play—a consortium of 44 domestic firms, including SoftBank, Sony, NEC, and Honda, dedicated to building independent AI for robots, vehicles, and factory floors. Tokyo has committed up to 1 trillion yen ($6.2 billion) over five years to this endeavor. NVIDIA is pivotal to this, constructing a massive “Vera Rubin AI factory” in Japan, slated for 2028, which will serve as the hardware backbone for Noetra’s ambitious physical AI models.
Additionally, Huang solidified alliances with Japan’s leading robotics and manufacturing players under the Cosmos initiative. Companies like Fanuc, Yaskawa, Kawasaki Heavy, and Fujitsu have committed to building on NVIDIA’s Cosmos models, an open-model effort designed for physical AI. NVIDIA further bolstered this commitment by unveiling Cosmos 3 Edge, a version of the model optimized to run on its Jetson Thor chips, directly within the machines themselves. This strategic move aims to empower Japan’s manufacturing base to reinvent intelligent industries, reinforcing Huang’s belief that Japan, as the inventor of modern manufacturing, is uniquely positioned to lead this transformation. The collaboration extends to Toyota, which is integrating NVIDIA’s Drive platform into its next-generation vehicles and leveraging NVIDIA’s technology for manufacturing simulations and advanced driver assistance systems.Contabo VPS
Source: TechCrunch
What to Read Next
- China’s Open-Weight AI Strategy in 2026: Why Kimi K3, Qwen 3.8, and DeepSeek Are Winning Globally
- Top 5 Free AI Tools in 2026 (No Credit Card, No Catch)
- Is Claude Pro Worth It in 2026? Honest Review After 3 Months
- Is Claude Pro Worth It in 2026? Honest Review After 3 Months
- Browse all AI Stack Digest articles
Bookmark aistackdigest.com for daily AI tools, reviews, and workflow guides.
This article was produced with the assistance of AI tools and reviewed by the AIStackDigest editorial team.