Meta Releases 4000 Billion Parameter LLaMA‑3‑Vision Model Free to Open‑Source Community

Meta announces free open‑source release of a 4 000‑billion‑parameter multimodal LLaMA‑3‑Vision model, enabling image and text processing, marking one of the largest parameter counts ever made publicly available. The model supports zero‑shot vision‑language tasks and is released under a permissive license for researchers and developers worldwide.

Hugging Face now hosts the Llama‑3.2‑1B model, a compact 1‑billion‑parameter version of Meta’s LLaMA family, providing developers easy API access and integration through its model hub. It offers text generation capabilities and can be fine‑tuned for niche applications, reinforcing the trend toward community‑driven AI development.

Open source AI advocates cite Meta’s release as a catalyst for democratizing large‑scale models, arguing that free access lowers entry barriers for academia and startups, potentially accelerating innovation in vision‑language research. They also warn that responsible use guidelines are essential to prevent misuse.

Industry analysts predict that the open‑source availability of such a massive model could spur competition, prompting other firms to release comparable high‑parameter models, while cloud providers may offer hosted inference services to meet demand.