Dec 6, 2024 at 8:21 PM Meta Unveils Llama 3.3 70B, its new cost-effective, high-performance open-source AI Model
Meta has introduced Llama 3.3 70B, the latest model in its Meta Llama series of generative AI models. This new model is designed to offer a cost-effective solution for AI inference, with Meta stating it operates at a fraction of the cost compared to other closed-source models.
The enhancements in Llama 3.3 are attributed to a new alignment process and advancements in online reinforcement learning techniques. Despite its smaller size, the 70B model rivals the performance of the previous Llama 3.1 405B model and is capable of running locally on standard developer workstations.
Llama 3.3 is open-source and currently available on Meta, GitHub and Hugging Face, with plans for deployment on Microsoft Azure and other major cloud platforms.
Comments
Gu
To their credit, Meta is one of the few companies that are really pushing frugal AI models to be available cheaply and locally. Certainly because Meta has no real plan to make real money from their models, and doesn't need that money any time soon. Not that I want defend Meta in any way but it seems that all other companies don't care about AI being sustainable and dive into this giant abyss of money, that make this technology one of most expensive ever, just hopping to find some sort of AGI that will never exist using LLM.