MLX Optimization Challenge Speeds Up Qwen 27B on Apple Silicon by Over 150%

A community MLX optimization challenge has boosted Qwen 3.8 27B performance on Apple Silicon by over 150% versus baseline, with MLX Fast achieving a 153% gain in under 16 hours and running 2.5x faster than native MTP decoding, overturning the notion that Macs can't handle dense models.

2026-08-16 ~ 2026-08-17 · 4 related posts

• Qwen 3.8 runs 151% faster on Apple Silicon via community challenge — gajesh · 2026-08-16
• Challenge boosts Qwen 3.8 speed by 152% on Mac — gajesh · 2026-08-16
• Qwen 27B runs 170% faster on Apple Silicon, challenging dense model bottlenecks — gajesh · 2026-08-16
• MLX Fast speeds up Qwen 3.8 27B by 153% on Apple Silicon — alexcovo_eth · 2026-08-17