@meta
  v: 1
  route: /model/llama-3-3-70b
  generated: 2026-06-10T08:51:35.863Z

@intent
  purpose:    Show what hardware can run Llama 3.3 70B.
  audience:   ai-engineer, self-hoster, model-evaluator
  capability: inspect_model_specs, find_compatible_gpus, compare_similar_models, open_huggingface

@state
  slug: llama-3-3-70b
  name: Llama 3.3 70B
  hf_repo: meta-llama/Llama-3.3-70B-Instruct
  params_b: 70.6
  active_params_b: ~
  family: Meta
  type: dense
  context_k: 128
  fp16_gb: 141.2
  summary: Meta's late-2024 70B. Closes the gap to GPT-4o-mini at the same VRAM footprint as 3.1 70B.
  smallest_fitting_gpu_q4: m2-max-64
  smallest_fitting_gpu_fp16: h100-nvl-2x

@actions
  - id: open_huggingface
    method: GET
    href: https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct
  - id: view_compatible_gpus
    method: GET
    href: /model/llama-3-3-70b
  - id: compare_similar
    method: GET
    href: /model/qwen-2-5-72b
  - id: view_calculator
    method: GET
    href: /#calculator

@context
  > Meta's late-2024 70B. Closes the gap to GPT-4o-mini at the same VRAM footprint as 3.1 70B.

@nav
  self:      /model/llama-3-3-70b
  parents:   [/]
  peers:     [/model/qwen-2-5-72b, /model/mixtral-8x7b, /model/command-r-plus, /model/qwen-3-6-35b-a3b]
  drilldown: /the-math