LLMCheck: Apple M5 Ultra (96GB), MLX, Qwen3-Coder-Next 80B Q4_K_M decode
72.0 tok/s (decode, per-request)
Fields
id
benchmarks/llmcheck-m5-ultra-qwen3-coder-next-mlx-decodeMetric
decode_tok_s
Value
72
Unit
tok/s (decode, per-request)
Model
Qwen3-Coder-Next
Methodology
THIRD-PARTY MEASURED (LLMCheck open dataset, dated 2026-08). M5 Ultra 96GB, MLX, Qwen3-Coder-Next (80B MoE) Q4_K_M, per-request.
Measured by
third_party
Reproducible
yes
notes
An 80B MoE decodes at 72 tok/s — well into interactive territory on unified memory.
Notes
An 80B MoE decodes at 72 tok/s — well into interactive territory on unified memory.