[Education] Offline benchmark performance of Qwen3-0.6B on MLX (CPU) and Modal (GPU)

[lamng3]

Offline benchmark performance of Qwen3-0.6B on MLX (CPU) and Modal (GPU)

This PR compares the performance of the Qwen3-0.6B model running on two different platforms: GPU acceleration via Modal (A10G) and CPU inference using MLX (Apple Silicon M1).

Modal

• GPU: A10G
• Model: Qwen/Qwen3-0.6B
• Total: 133966tok
• Time: 44.10s
• Throughput: 3037.56tok/s

MLX Perf

• CPU: Apple Silicon M1
• Model: mlx-community/Qwen3-0.6B-4bit
• Total: 140435tok
• Time: 1017.77s
• Throughput: 137.98tok/s