Llama 3.2 1B

Llama 3.2 1B is Meta’s very small autoregressive language model for text generation and understanding. It is designed for cases where a compact footprint matters, such as constrained or edge-oriented deployments, though the matrix only claims supported hardware rather than a particular product use. The 1B name means about one billion parameters. It is text-only; the vision-capable member of Llama 3.2 in this deck is the separate 11B Vision model. TT-Metal lists Llama 3.2 1B on a single n150 Wormhole device with batch 32. No tensor- or data-parallel factor appears in that row, making it one of the simplest hardware scales in the language-model section.