Meta AI adds 405B powerhouse and AI selfies
Meta just rolled out a major upgrade to Meta AI, adding preview access to its flagship Llama 3.1 405B model alongside a new "Imagine me" stylized selfie feature. Users can now generate custom scenes starring themselves, though the 405B model requires manual switching and has weekly limits.
Meta is rolling out its biggest Meta AI update yet, bringing its new heavyweight Llama 3.1 405B model and personalized generative selfies right into your social feeds.
Meta introduced a series of major upgrades across Facebook, Instagram, Messenger, and the web. The headline feature is access to Llama 3.1 405B, which Meta claims can handle complex tasks that earlier iterations notoriously bungled, including math homework, scientific explanations, and code debugging.
Flagship intelligence behind a velvet rope
Early versions of Meta AI faced widespread criticism for struggling with basic web searches, fact-checking, and simple queries like looking up airfares. The new 405B model aims to change that, but it comes with strings attached.
Users have to manually toggle to the 405B model, which Meta currently labels as a preview. Access is also restricted by a weekly query limit, after which the assistant quietly downgrades you to the smaller Llama 3.1 70B model.
Alongside the text upgrade, Meta launched a new generative image tool called Imagine Yourself. In beta, users can type "Imagine me" followed by a prompt like "surfing" or "on a beach vacation" to generate stylized portraits based on their own photos.
Meta did not detail what training data went into Imagine Yourself. However, its terms of service explicitly permit training on public posts and photos across its platforms, a policy that continues to irritate privacy-conscious users navigating its convoluted opt-out process.
Social tweaks and Quest vision
Meta is also leaning hard into distribution. New prompt-based editing lets users replace objects in generated images, with a dedicated Edit with AI button slated for next month and quick-share buttons rolling out to feeds, stories, and comments.
Hardware is getting looped in as well. Starting next month in the U.S. and Canada, Meta AI will replace Quest VR voice commands in an experimental mode. With passthrough active, wearers can ask the assistant about physical objects around them, such as asking for wardrobe advice while holding up a pair of shorts.
Meta AI is now available in 22 countries, adding markets like Argentina, Chile, Colombia, Mexico, and Cameroon, along with support for French, German, Hindi, Italian, Portuguese, and Spanish.
Why it matters
Meta is doing what it does best: turning cutting-edge AI into casual, viral consumer toys. Placing personalized AI selfies and a frontier-class model directly into apps billions of people already use lowers the discovery barrier to zero, putting severe pressure on standalone chatbots.
However, rationing the 405B model with weekly caps shows just how brutal the compute economics remain. If regular users constantly bump into limits and fall back to smaller models, the perceived leap in intelligence might feel hollow.
We also see the Quest rollout as a calculated dry run. Integrating multimodal vision into head-worn hardware is the real playground for Meta, serving as a stepping stone toward lightweight smart glasses that understand what you are looking at in the physical world.
Sources (1)
- techcrunch.com(opens in a new tab)Originaltechcrunch.com