verifiedmedium · 0.80updated 2026-10-15

LLMCheck: Apple M5 Ultra (256GB), MLX, Qwen3-235B-A22B Q4_K_M decode

37.0 tok/s (decode, per-request)

Fields

benchmarks/llmcheck-m5-ultra-qwen3-235b-mlx-decode

Metric

decode_tok_s

Value

Unit

tok/s (decode, per-request)

Model

Qwen3-235B-A22B

Methodology

THIRD-PARTY MEASURED (LLMCheck open dataset, dated 2026-08). M5 Ultra 256GB, MLX, Qwen3-235B-A22B Q4_K_M, per-request.

Measured by

third_party

Reproducible

yes

notes

37 tok/s on a 235B MoE is the Ultra's reason to exist: double the Max's 18 tok/s on the same model, comfortably interactive.

37 tok/s on a 235B MoE is the Ultra's reason to exist: double the Max's 18 tok/s on the same model, comfortably interactive.