News at a Glance
Meta held its annual Connect conference on September 23, 2026, in Menlo Park, where CEO Mark Zuckerberg delivered a keynote outlining the company's vision to build a future for everyone. The event showcased Meta's latest AI models and hardware, including the Quest Pro 2 headset and upgraded Ray-Ban smart glasses, posit
What Happened
Meta hosted Connect 2026 on September 23, 2026, at its Menlo Park headquarters, with CEO Mark Zuckerberg delivering a 90-minute keynote. Zuckerberg reiterated Meta's mission to build a future for everyone, highlighting progress in AI and smart glasses. The event featured the new Quest Pro 2 mixed reality headset, powered by Qualcomm Snapdragon XR3+ Gen 2, with dual 4K Micro-OLED displays at 120Hz and a 15% weight reduction. Meta also unveiled upgraded Ray-Ban smart glasses with an enhanced Meta AI assistant supporting real-time translation and scene understanding. Zuckerberg announced Llama 4 Ultra, a 2-trillion-parameter model surpassing GPT-4o and Gemini 1.5 Pro on seve
Background
Connect 2026 comes at a critical juncture for Meta. In 2025, Meta invested over $40 billion in AI, facing stiff competition from OpenAI and Google. Zuckerberg has repeatedly emphasized AI and smart glasses as Meta's next growth curve. At Connect 2024, Meta launched Quest 3S and Ray-Ban Meta glasses, receiving positive market response, though the mixed reality business remains unprofitable. The release of Llama 4 Ultra aims to solidify Meta's leadership in open-source AI, while AI Studio seeks to attract developer ecosystem. Industry analysts believe Meta
Deep Dive
Liu Gong believes the release of Llama 4 Ultra represents a major bet in open-source AI. With 2 trillion parameters, it competes with closed-source models in complex reasoning, but Liu Gong notes challenges in training cost and inference efficiency. Compared to GPT-4o, Llama 4 Ultra improves MMLU by about 8 percentage points, yet inference costs remain 30% higher. Liu Gong predicts optimization through quantization and distillation, but short-term cost-effectiveness still favors closed-source solutions. Liu Gong points out Meta's hardware strategy is shifting from selling devices to building an ecosystem. Quest Pro 2 is priced at $999, down $200 from the previous generation, but Meta's focus is on developer stickiness through AI Studio. Liu Gong analyzes that Meta is replicating Apple's model through hardware+AI+app store integration. However, Liu Gong cautions that Meta's developer revenue-sharing policies remain conservative, risking developer diversion to Apple's visionOS and Google's Android XR. Liu Gong believes this launch intensifies the AI hardware race. Meta's continued investment in open-source models differentiates it from Google's Gemini and OpenAI's GPT. Liu Gong predicts smart glasses will become a key AI interaction entry point within 12 months, and Meta's first-mover advantage could translate into meaningful market share. However, Liu Gong warns that Meta's historical controversies over privacy and content moderation may hinder large-scale AI deployment.
Data Points
- Llama 4 Ultra has 2 trillion parameters, roughly 4x larger than Llama 3.1's 405 billion; it scores 89.7% on MMLU, 3.3 points higher than GPT-4o's 86.4%. Quest Pro 2 is priced at $999, down 33% from Quest Pro's $1,499; it weighs 480g, 15% lighter than the 565g predecessor. Ray-Ban Meta glasses cumulative sales exceed 3 million units, up 120% year-over-year. Meta AI assistant has 120 million daily active users, up 50% since early 2025.
Impact & Outlook
For the industry, within 3 months, Llama 4 Ultra will intensify competition in open-source AI reasoning, prompting Google and OpenAI to accelerate next-gen model releases. Within 1 year, if AI Studio attracts developers, Meta could form a third major AI app ecosystem alongside Apple and Google. For consumers, within 3 months, Quest Pro 2 and upgraded Ray-Ban glasses offer more immersive mixed reality and smarter AI assistance; within 1 year, smart glasses may become a daily AI interaction entry point, though users should monitor privacy and content moder
Angles to Explore
Points to consider
- Can Llama 4 Ultra's open-source strategy truly challenge closed-source commercial advantages?
- Will Meta's AI Studio developer ecosystem break through amid Apple and Google competition?
- How to balance privacy and security risks as smart glasses become AI interaction entry points?
Source & Original
This item comes from The Verge AI (published 2026-09-23 22:30:00). Wenfeng Radar provides a Chinese/English summary and commentary on overseas AI developments; the original content is owned by its author.
Daily aggregation of overseas AI developments with in-depth analysis. Bookmark us to never miss an important signal; back to homepage.