Llama 3.1 8B

Llama 3.1 8B is Meta’s smaller general-purpose autoregressive text language model from the Llama 3.1 family. At roughly eight billion parameters, it is far more compact than Llama 3.1 70B while serving the same broad class of generation and understanding tasks. It is text-only, unlike Llama 3.2 11B Vision. TT-Metal’s matrix gives this one model several deployments: single p100 and p150 Blackhole cards at batch 32, two p150 Blackhole cards with DP=2 at batch 64, and a single n150 Wormhole at batch 32. Those rows show portability across two accelerator generations and optional data-parallel replication; they do not denote four separate Llama models.