MLX-LM on 64GB Apple M4 Max: Qwen3-30B-A3B-Instruct-2507 4-bit prompt processing at 2048 tokens
1753.9 tok/s (prompt processing of a 2048-token prompt, batch 1)
Fields
id
benchmarks/mlx-m4-max-qwen3-30b-a3b-q4-prefill-tok-sMetric
prefill_tok_s
Value
1,753.9
Unit
tok/s (prompt processing of a 2048-token prompt, batch 1)
Model
Qwen3-30B-A3B-Instruct-2507 (MLX 4-bit, mlx-community repo)
Methodology
mlx_lm.benchmark --model mlx-community/Qwen3-30B-A3B-Instruct-2507-4bit -p 2048 -g 128. Batch size 1, 2048 prompt tokens, 128 generated tokens. Hardware: Apple M4 Max 64GB unified memory. Software: mlx 0.29.2.dev20251008+85a8824a8, mlx-lm 0.28.2, macOS 26.1. Peak memory: 18.20 GB. MMLU Pro: 70.71.
Measured by
vendor
Reproducible
yes
notes
Official benchmark published by Apple MLX maintainers in mlx_lm/BENCHMARKS.md. Shows prompt processing throughput of 1753.90 tok/s for Qwen3-30B-A3B at 4-bit, paired with 113.33 tok/s generation throughput on the 64GB M4 Max package.
Notes
Official benchmark published by Apple MLX maintainers in mlx_lm/BENCHMARKS.md. Shows prompt processing throughput of 1753.90 tok/s for Qwen3-30B-A3B at 4-bit, paired with 113.33 tok/s generation throughput on the 64GB M4 Max package.