mlboydaisuke commited on
Commit
dcda323
·
verified ·
1 Parent(s): a967f69

card: add the raw-MLX arm — Core AI ties MLX, both beat the ExecuTorch build by 14%

Browse files
Files changed (1) hide show
  1. README.md +24 -0
README.md CHANGED
@@ -80,6 +80,30 @@ code falls to 1.40× and free chat to **30.10 tok/s, below their number**; tool
80
  matters more than acceptance does: verify cost here is a staircase (S ≤ 3 free, S = 4…8 ~1.47×,
81
  S ≥ 9 ~2.3×), so K=8 makes free chat 5% *slower* while K=2 makes it 34% faster.
82
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83
  **Not claimed here:** Meta's published quality figure (1.0% degradation across 15 benchmarks
84
  for the 17G quant) has no matched counterpart; this port has a token-exact gate and read
85
  generations, not a benchmark suite.
 
80
  matters more than acceptance does: verify cost here is a staircase (S ≤ 3 free, S = 4…8 ~1.47×,
81
  S ≥ 9 ~2.3×), so K=8 makes free chat 5% *slower* while K=2 makes it 34% faster.
82
 
83
+ ### Raw MLX is the third arm, and it changes what the ExecuTorch result means
84
+
85
+ ExecuTorch's `metal` backend is MLX-native, so beating it could mean beating MLX or beating
86
+ the wrapper around MLX. Only raw MLX separates the two. Same machine, same prompts, greedy,
87
+ 192 tokens, interleaved CA/ET/MLX with a 45 s cooldown between every run:
88
+
89
+ | | p1 r1 | p1 r2 | p2 r1 | p2 r2 | mean |
90
+ | --- | ---: | ---: | ---: | ---: | ---: |
91
+ | **Core AI** `int4hu`, 16.35 GB | 27.5 | 27.1 | 27.5 | 27.6 | **27.43** |
92
+ | **MLX** `mlx-community/…-4bit`, 18 GB | 27.18 | 27.38 | 27.50 | 27.37 | **27.36** |
93
+ | **ExecuTorch** `k-quant-17G…metal`, 17.9 GB | 24.0 | 23.8 | 24.1 | 24.1 | **24.00** |
94
+
95
+ **Core AI and raw MLX are indistinguishable (+0.3%). Both beat Meta's own build by ~14%.**
96
+
97
+ So the honest reading is not "Core AI is fast here" — it is that **Meta's shipped on-device
98
+ artifact leaves ~14% on the table against the runtime it is built on**. Core AI matching MLX
99
+ at 27.9 B dense is what [`coreai-vs-mlx-speed.md`](https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/coreai-vs-mlx-speed.md)
100
+ already predicts: Core AI ≥ MLX on small dense, converging to a tie as the model grows and
101
+ MLX's 4-bit byte advantage cashes in. This is the largest dense point on that curve so far,
102
+ and it lands on the tie.
103
+
104
+ Prompt processing is not matched here and no claim is made from it (Core AI 269 tok/s at 512
105
+ prompt tokens, MLX 128–136 at 77–81 — different lengths, different batching).
106
+
107
  **Not claimed here:** Meta's published quality figure (1.0% degradation across 15 benchmarks
108
  for the 17G quant) has no matched counterpart; this port has a token-exact gate and read
109
  generations, not a benchmark suite.