Member

quietmodel

@quietmodel · joined 13 Sept 2026 · 1 public runs · 1 public reviews

Sample content. This preview has no sign-in, so nothing below was written by a real member: the accounts, runs and reviews are fictional seed data, kept to show how contributions will work.

Benchmark runs

54.2 tok/s generation throughput

Qwen3 30B-A3B MLX 4-bit

on Apple M4 Pro (20-core GPU) (member system, 48 GB unified) · MLX-LM 0.24.0 · metal

@quietmodel · 3 May 2025 unverified

“Thinking disabled.”

Full environment and measurements

Benchmark: Interactive throughput
Generation throughput: 54.2 tok/s
Peak memory: 17.8 GB
Prompt throughput: 520 tok/s
Time to first token: 610 ms
Artifact: Qwen3 30B-A3B MLX 4-bit (MLX Community)
Context: 8K
OS: macOS 15.4

Reviews

MoE makes a 48 GB Mac feel fast

@quietmodel on Qwen3 30B-A3B · 13 Sept 2026

Runs at interactive speeds in MLX 4-bit. Thinking mode helps on multi-step problems; turn it off for quick chat.

Coding4/5

Reasoning4/5

Agents & tool use4/5

Hardware efficiency5/5