$ onpar learn basics 1 --no-quiz


────────────────────────────────────────────────────────────
  [basics.1]  What a model is in bytes (weights)

  lesson basics·1 — what a model is in bytes (weights)

  A language model is a large array of learned numbers — the weights.
  Everything else (layers, attention heads, MLP) is how they're organised.

  Llama 3.3 70B has 70.6B parameters.
  At q8 (8 bits per weight):

    weight bytes = 70.6B × 10⁹ × 8 bits ÷ 8
                 = 65.8 GB

  Your M4 Max has 128 GB total.
  onpar counts 96 GB usable (75%, a calibration knob).

  Just the weights consume 68% of that budget.

  answer: 6.5 GB  (7B × 10⁹ × 8 bits ÷ 8 = 6.5 GB)



# recorded 2026-11-03 on M4 Max · 128 GB · onpar 1.5.2
