Anyone can download and run Meta's new advanced AI model
Meta has unveiled Llama 3.1, the largest open-source artificial intelligence (AI) model to date. This new model surpasses top private models such as OpenAI's GPT-4o and Anthropic's Claude 3.5 Sonnet in several benchmarks. "Llama 3.1 405B is the first openly available model that rivals the top AI models when it comes to state-of-the-art capabilities in general knowledge, steerability, math, tool use, and multilingual translation," said Meta. Company CEO Mark Zuckerberg called it "the first frontier-level open source AI model."
Llama 3.1: A leap forward in AI complexity and performance
Llama 3.1 is a significant advancement over its predecessors, boasting an impressive 405 billion parameters. The model was trained using more than 16,000 of NVIDIA's high-end H100 GPUs. Although Meta has not revealed the exact cost of developing Llama 3.1, it is estimated to be in the hundreds of millions of dollars based solely on the price of the NVIDIA chips used.
Meta's open-source vision: A game-changer for AI industry
It is perhaps the first time anyone can download an advanced large language model (LLM) for free and run it on their own hardware. Per Meta, it can run on a "single server node," which isn't desktop PC-grade equipment. Zuckerberg believes that "open-source AI models will surpass proprietary models in terms of improvement speed and prevalence." He predicts that "the Llama 3.1 release will be an inflection point in the industry where most developers begin to primarily use open source."
Meta collaborates with tech giants for Llama 3.1 deployment
To facilitate the deployment of Llama 3.1, Meta is partnering with over two dozen companies, including Microsoft, Amazon, Google, NVIDIA, and Databricks. The company claims that running Llama 3.1 costs approximately half as much as running OpenAI's GPT-4o in production. Additionally, Meta is releasing the model weights so companies can train it on custom data and adjust it to their needs.
Llama 3.1: A cost-effective teaching tool for smaller models
Meta used synthetic data, or data generated by a model rather than humans, to enhance the smaller 70 billion and 8 billion versions of Llama 3.1. Ahmad Al-Dahle, Meta's VP of generative AI, anticipates that Llama 3.1 will be popular with developers as "a teacher for smaller models that are then deployed" in a "more cost-effective way." For the first time, Meta's red teaming of Llama 3.1 included potential cybersecurity and biochemical use cases.
Meta AI assistant expands reach, adds new features
Meta's AI assistant, which uses Llama, is available across Instagram, Facebook, and WhatsApp. Initially accessible through WhatsApp and the Meta AI website in the US, it will soon be available on Instagram and Facebook. The assistant will support new languages including French, German, Hindi, Italian, and Spanish. A new "Imagine Me" feature allows users to insert their likeness into images generated by the assistant by scanning their face through their phone's camera.
Meta AI to replace voice command interface on Quest headset
In the coming weeks, Meta AI will replace the voice command interface on the Quest headset. Users will be able to use Meta AI on the Quest to identify and learn about what they're looking at while in passthrough mode. Despite Zuckerberg's claim that Meta AI will become the most-used chatbot by year-end, surpassing ChatGPT's over 100 million users, Meta has not yet released any usage numbers for its assistant.
