Meta has officially released the Muse Glimmer 30B model, a 30-billion-parameter AI model designed specifically for local AI agents and consumer gadgets. According to Business Standard, the model is optimized for local AI agents that can work continuously without depending entirely on cloud infrastructure or internet connection. CEO Mark Zuckerberg confirmed the release in a post on X, stating 'Today we're also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally.' The model can run on a Mac or PC with a single consumer GPU, addressing the growing demand for AI systems that operate directly on people's devices. As per Business Standard, the company has released the model weights under the Apache 2.0 licence and developers can download them from Hugging Face. The model is quantized to fit under 20 GB and requires minimal hardware specifications, making it accessible to consumer devices.
CEO Mark Zuckerberg outlined Meta's ambitious vision for universal superintelligence access in a comprehensive 6,500-word essay titled 'The Future is for Everyone', as reported by Mint. The company plans to offer highly capable personal AI agents that can work across areas such as health, careers, finances, relationships and home management. Zuckerberg emphasized that everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about. The personal agents will have strong privacy and security options and can be accessed through different devices, including smart glasses. Zuckerberg pledged free or affordable access to a personal AI agent for every user, with users seeking additional computing resources able to pay through a dynamic auction mechanism. In his latest essay, Zuckerberg championed spreading AI rather than leaving it in the hands of a few, stating 'The notion AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic.' He wrote that 'Everyone will soon have invention superpowers' and noted that his 8-year-old daughter can already code her ideas and quickly produce videos.
Muse Glimmer demonstrates significant capabilities beyond traditional AI models, with complete end-to-end task handling including information searching, code writing and debugging, and multi-turn request resolution. According to Business Standard, the model can handle function calling, multistep reasoning, and coding while maintaining reasoning across complex tasks. It supports more than 100 languages and includes a dedicated perception encoder for processing text and images. The model features error recovery capabilities - if a tool call fails or returns unexpected results, it can identify the error and retry instead of stopping. Muse Glimmer is compatible with OpenClaw and other agentic orchestration frameworks and supports different levels of reasoning to balance response quality with speed. The company has developed integrations with llama.cpp, MLX and ExecuTorch that will be available in the coming days, with support for vLLM and SGLang for serving at scale.
CEO Mark Zuckerberg released a 14-page essay supporting the open-source ecosystem over closed-source competitors, as reported by The Economic Times and Business Standard. The essay was released alongside Meta's new open-weight AI model, with Zuckerberg advocating for lower US barriers for open-source AI to better compete with Chinese rivals. According to Business Standard, Zuckerberg called for US policy to reduce additional friction if American open-source models are to lead over time, including on data use and distillation methods. He criticized that 'Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data'. Zuckerberg announced that Meta would implement a governance structure to give its independent directors the power to approve the safety criteria for releasing models. The company plans to release the weights of Muse Spark 1.2, its most advanced model, built by a costly superintelligence team formed last year. Meta will continue supporting open-source AI and plans to resume releasing some open-source models soon.
The AI agent revolution is creating unprecedented security challenges as organizations rapidly adopt autonomous systems. Recent research reveals that 88% of organizations report suspected or confirmed AI agent security incidents, yet only 22% need new approval processes, highlighting a critical gap in enterprise security architecture. Major technology companies are responding with comprehensive security frameworks. Cisco has launched Cisco Secure AI Factory with NVIDIA, featuring 1024 NVIDIA Blackwell Ultra GPUs and powered by Cisco's UCS servers, delivering breakthrough performance for complex AI tasks. Palo Alto Networks has advanced its Prisma AIRS 3.0 platform with end-to-end discovery, risk assessment, and runtime protection capabilities. Okta is launching Okta for AI Agents platform on April 30, 2026, providing organizations with centralized control over agent deployment, access, and governance across hybrid environments.
