Mark Zuckerberg took to the stage at Meta’s Menlo Park campus on Wednesday to unveil three distinct smart glasses models, signalling the company’s most aggressive push yet to turn eyewear into the primary interface for generative artificial intelligence. The announcements, made during the keynote of the annual Connect developer conference, position Meta squarely against Apple and a wave of startups racing to define the post-smartphone era.
The Line-Up: From Everyday to Immersive
The headliner is the Ray-Ban Meta Skyler, a refreshed take on the existing Wayfarer partnership. It retains the familiar aesthetic but packs a upgraded 12-megapixel ultra-wide camera, five-microphone spatial audio capture, and the new Snapdragon AR1 Gen 1 platform. Crucially, it ships with Meta AI with Vision enabled by default in the US and Canada, allowing wearers to identify landmarks, translate text, or generate captions for Instagram simply by saying “Hey Meta, look at this.”
Sitting a rung above is the Meta Orion — the device formerly whispered about as “Project Nazare.” This is not a consumer product, nor is it a traditional VR headset. Orion is a chunky, magnesium-framed prototype that beams full-colour holograms onto silicon-carbide waveguides using micro-LED projectors. Zuckerberg called it “the most advanced AR glasses ever made,” boasting a 70-degree field of view, eye-and-hand tracking, and a wireless compute puck the size of a vape pen. Meta plans to ship roughly 1,000 units to internal developers and select partners next year; a commercial version remains “a few years out.”
Rounding out the trio is a budget-friendly, display-less model codenamed “Astra” internally, aimed at emerging markets and enterprise fleets. It strips out the camera and speakers, relying entirely on bone-conduction audio and an always-on, low-power large language model running locally on a new Meta-designed chip. Battery life is quoted at 18 hours. Pricing starts at $199.
The AI Layer: Context Over Commands
What binds the three devices is Meta AI, now powered by the Llama 3.2 multimodal family. Unlike the voice assistants of a decade ago, this system is designed for ambient interaction. On the Skyler, the AI sees what the user sees, enabling real-time object recognition and live translation across 36 languages. On Orion, the same models drive spatial computing — anchoring digital windows to physical walls, recognising gestures, and surfacing contextual information without a prompt.
“We’re moving from ‘hey assistant, do this’ to ‘glasses, I’m looking at a menu in Italian, translate it and tell me what’s vegetarian’,” Zuckerberg told the crowd. “That requires the model to understand the world with you, not just for you.”
Meta also announced a new Developer SDK for Wearable AI, opening access to the on-device vision encoder, the spatial anchoring API, and a privacy-first framework that processes sensitive visual data locally unless the user explicitly consents to cloud offload.
Privacy, Power, and the Platform Play
The privacy architecture drew immediate scrutiny. Meta insists that on the Skyler and Astra, raw video never leaves the device; only anonymised embeddings are uploaded for model improvement, and users can toggle a physical “capture” LED that cannot be disabled in software. Orion, with its always-on cameras for hand tracking, adopts a similar local-first approach but adds a hardware shutter for the outward-facing sensors.
Power remains the stubborn physics problem. The Skyler manages four hours of mixed use — down from six on the previous generation — because the AR1 chip runs the vision transformer continuously. The Orion compute puck houses a custom 5 nm SoC and a 3,000 mAh cell, good for roughly two hours of holographic rendering. Meta says it is working with EssilorLuxottica on frame-integrated batteries and with Amprius on silicon-anode cells that could double density by 2026.
Why it Matters
Meta is no longer treating smart glasses as a peripheral accessory; it is recasting them as the primary compute surface for the AI age. By launching three form factors simultaneously — consumer fashion, developer-grade AR, and ultra-low-cost utility — the company is flooding the zone to lock in developers, normalise face-worn cameras, and gather the real-world multimodal data its models crave. If even one of these sticks, the smartphone’s grip on our attention begins to loosen, and the battle for the next platform shifts from the pocket to the bridge of the nose.
