Topic: open-weight ai models
-
Meta's new open models signal another AI strategy reset
Meta is pivoting back to open-weight AI models, releasing the compact, locally runnable Muse Glimmer (30B parameters, 128K context, Apache 2.0) and pledging to open the weights of its larger Muse Spark 1.2 model soon. CEO Mark Zuckerberg published a 6,000-word essay criticizing proprietary rivals...
Read More » -
Meta's Glimmer AI hints at Zuckerberg's personal intelligence push
Meta released Muse Glimmer, an open-weight 30-billion parameter model under Apache 2.0, designed to run multi-step AI agents locally on consumer devices with a single GPU, supporting text, images, and 100+ languages. Glimmer enables privacy-conscious, always-on personal tasks like schedule manage...
Read More » -
Meta Invests $21B More in CoreWeave for AI Cloud
Meta has expanded its partnership with CoreWeave, committing an additional $21 billion for AI inference capacity from 2027-2032, bringing their total agreement to about $35 billion and enabling early use of Nvidia's Vera Rubin platform. CoreWeave, originally a crypto mining firm, pivoted to GPU c...
Read More » -
PrismML Deploys Tiny LLMs on Qualcomm Smart Glasses
PrismML has successfully ported its specialized tiny language models to smart glasses powered by Qualcomm’s Snapdragon chips, marking a significant shift toward on-device AI processing. The company’s 1-bit Bonsai LLM operates locally on the AR1 platform, reducing latency and privacy concerns by e...
Read More » -
Why China’s Top AI Experts Are Also Worried
The primary message from a major AI conference in Beijing is that the US and China must set aside their fierce rivalry to address the serious cybersecurity and systemic risks posed by frontier AI, as cooperation is essential for global safety. Experts argue that international collaboration on AI ...
Read More » -
Meta follows SpaceX in monetizing excess AI compute power
Meta is preparing to launch a cloud infrastructure business selling AI compute power and model access, directly competing with Amazon Web Services, Google Cloud, and Microsoft Azure. The company has committed $182.9 billion to AI infrastructure, including massive data centers in Louisiana and Ohi...
Read More » -
Why AI companies are obsessed with forward-deployed engineers
Only about 2,000 engineers in the U.S. have the rare combination of expertise needed to deliver measurable ROI on enterprise AI investments, while demand for forward-deployed engineers (FDEs) is projected to surge by 2,100% by year-end. Enterprises are rapidly shifting from pilot projects to full...
Read More » -
DeepSeek-V3.1: China's Record-Breaking AI Model Unveiled
China has released DeepSeek-V3.1, a powerful new AI language model with 685 billion parameters, showcasing its growing competitiveness in the global AI landscape. The model features a dual-mode capability for reasoning and efficiency, supports a 128,000 token context, and is available under an op...
Read More »