Taiwan-based Wistron has launched a 324,000-square-foot factory in Texas to produce NVIDIA GB300 Grace Blackwell Ultra and Vera Rubin superchips. The $700 million investment has created over 500 jobs, with plans to hire 1,000 employees by year-end. The project is part of NVIDIA's commitment to building AI infrastructure in the United States.
NVIDIA
Poolside released Laguna S 2.1, an open-weight model with 11.8 billion parameters and 8 billion active per token, designed for agentic coding. Based on a Mixture-of-Experts (MoE) architecture, it supports up to 1 million tokens of context and can run on a single NVIDIA DGX Spark. Laguna S 2.1 ranks first among open models with known size on Terminal-Bench 2.1 (70.2%) and SWE-Bench Multilingual (78.5%), outperforming many larger competitors.
Poolside
DeepSeek
NVIDIA
Tencent
Moonshot AI
Alibaba/Qwen
Anthropic
The US Treasury Secretary stated that the administration is considering sanctions against Chinese AI models due to potential intellectual property theft. Nvidia CEO Jensen Huang in response urged not to fear Chinese open models, arguing that they only increase demand for chips and data centers.
NVIDIA
Moonshot AI
Anthropic
OpenAI
A developer built an MCP server to order groceries from Yandex Lavka through an AI assistant. This involved reverse-engineering the private API, tackling CSRF issues, cart race conditions, unavailable items from a different warehouse, and empty payment methods. The code is open-sourced under the MIT license.
Anthropic
At an investor meeting, DeepSeek founder Liang Wenfeng detailed the company's technological aspects and business strategy, including competition with OpenAI and partnerships with CATL. Topics included financing and positioning in the large language model market.
DeepSeek
Moonshot AI
OpenAI
NVIDIA
Cisco Foundation AI has released the Antares family of models (350M and 1B parameters), designed to localize vulnerabilities in source code. The models are available on Hugging Face under the Apache 2.0 license. They outperform many larger models in accuracy (File F1), while the cost per run on 500 tasks is less than $1.
Cisco Systems
IBM
Russia and China are discussing the possibility of creating a joint AI alliance, which would include building a network of data centers in Siberia and the Far East. The project aims to develop computing power for artificial intelligence.
Chinese AI developer Zhipu AI, whose value previously surged 24 times, lost 790 billion yuan in market capitalization. Experts attribute this to a revaluation of the large language model sector amid intensifying competition and slowing AI monetization.
A journalist combined four Mac Studio units with M3 Ultra into a cluster with a total memory of 1.5 TB using RDMA technology via Thunderbolt 5. The cluster ran massive AI models such as Kimi K2 Thinking (over 600 GB), which cannot be loaded on a single computer. The project was implemented using the open-source tool Exo 1.0.
Apple
NVIDIA
Exo
Moonshot AI
DeepSeek
Alibaba/Qwen