M6 family
M5 family
M4 family
M3 family
M2 family
M1 family
By Mac model
Head to head
- M1 vs M1 Pro
- M1 Pro vs M1 Max
- M1 Max vs M1 Ultra
- M1 vs M2
- M1 Pro vs M2 Pro
- M2 vs M2 Pro
- M1 Max vs M2 Max
- M2 Pro vs M2 Max
- M1 Ultra vs M2 Ultra
- M2 Max vs M2 Ultra
- M2 vs M3
- M2 Pro vs M3 Pro
- M3 vs M3 Pro
- M2 Max vs M3 Max
- M3 Pro vs M3 Max
- M2 Ultra vs M3 Ultra
- M3 Max vs M3 Ultra
- M1 vs M4
- M2 vs M4
- M3 vs M4
- M1 Pro vs M4 Pro
- M3 Pro vs M4 Pro
- M4 vs M4 Pro
- M1 Max vs M4 Max
- M2 Max vs M4 Max
- M2 Ultra vs M4 Max
- M3 Max vs M4 Max
- M3 Ultra vs M4 Max
- M4 Pro vs M4 Max
- M4 vs M5
- M4 Pro vs M5 Pro
- M5 vs M5 Pro
- M1 Max vs M5 Max
- M3 Ultra vs M5 Max
- M4 Pro vs M5 Max
- M4 Max vs M5 Max
- M5 Pro vs M5 Max
- M3 Ultra vs M5 Ultra
- M4 Max vs M5 Ultra
- M5 Max vs M5 Ultra
- M5 vs M6
Mac vs PC
Or start from a model: every model · compare
How it’s worked out
Memory: the model’s parameters at 8, 6, 4, 3 or 2 bits per weight plus MLX’s group scales (half a bit per weight), or its native format, with room for the context on top. By default macOS lets the GPU use about 75% of unified memory. Speed: memory bandwidth divided by the bytes read per token is the ceiling for one conversation; in llama.cpp’s public Apple Silicon benchmarks real runs reach about 45% to 80% of it, which is the range shown. These are estimates, not measurements: measure your own Mac on MLX Bench. Model figures from each official model card, chip figures from Apple’s Tech Specs. Checked 2026-10-19. Same data as mlx.app/compare.