
IQ2_XS: 9 tensors, 765.5M weights
  ggml native 74 B/block   shipped nibble 138 B   grid 82 B
  MiB served today    393.5   grid    233.8   saving    159.7   (ggml's own 211.0)
    (5120, 17408) x5
    (17408, 5120) x3
    (5120, 10240) x1

IQ2_S: 17 tensors, 2498.2M weights
  ggml native 82 B/block   shipped nibble 138 B   grid 82 B
  MiB served today   1284.3   grid    763.1   saving    521.2   (ggml's own 763.1)
    (5120, 17408) x9
    (17408, 5120) x4
    (5120, 6144) x2
    (5120, 248320) x1
    (5120, 1024) x1

IQ3_XXS: 78 tensors, 5421.1M weights
  ggml native 98 B/block   shipped nibble 134 B   grid 102 B
  MiB served today   2706.2   grid   2059.9   saving    646.2   (ggml's own 1979.1)
    (5120, 17408) x39
    (5120, 10240) x13
    (5120, 6144) x9
    (17408, 5120) x7
    (6144, 5120) x5
    (5120, 12288) x3
    (5120, 1024) x2

IQ1_M: 1 tensors, 89.1M weights
  ggml native 56 B/block   shipped nibble 138 B   grid 58 B
  MiB served today     45.8   grid     19.3   saving     26.6   (ggml's own 18.6)
    (5120, 17408) x1

IQ2_XXS: 5 tensors, 382.7M weights
  ggml native 66 B/block   shipped nibble 134 B   grid 70 B
  MiB served today    191.1   grid     99.8   saving     91.2   (ggml's own 94.1)
    (5120, 17408) x3
    (5120, 12288) x1
    (5120, 10240) x1
