LLMCheck: Apple M5 Max (128GB), MLX, Qwen3.5-35B-A3B Q4_K_M decode
52.0 tok/s (decode, per-request)
Fields
id
benchmarks/llmcheck-m5-max-qwen35-35b-a3b-mlx-decodeMetric
decode_tok_s
Value
52
Unit
tok/s (decode, per-request)
Model
Qwen3.5-35B-A3B
Methodology
THIRD-PARTY MEASURED (LLMCheck open dataset). M5 Max, MLX, Qwen3.5-35B-A3B (MoE, ~3B active) Q4_K_M, TTFT 0.9 s, per-request.
Measured by
third_party
Reproducible
yes
notes
The MoE point the user cares about: ~4.3x the dense-70B decode on the same chip. Qwen3.6-35B-A3B lands at 55 tok/s (separate record).
Notes
The MoE point the user cares about: ~4.3x the dense-70B decode on the same chip. Qwen3.6-35B-A3B lands at 55 tok/s (separate record).