Is the M5 faster than the M4 for LLMs?

The M5 reads memory about 27% faster than the M4 (153 vs 120 GB/s), so it writes about 27% faster with any model that fits both. Reading a long prompt depends on GPU compute instead, which this comparison doesn’t cover.

Which can run bigger models, the M4 or the M5?

Neither: both go up to 32 GB. At that size the M4 runs 27 and the M5 28 of 73 open models well; the faster one clears more of them at 10 tokens a second.

M4 vs M5 for local LLMs: memory, speed, which to buy

Side by side

	M4	M5
Memory sizes	16, 24, 32 GB	16, 24, 32 GB
Memory bandwidth	120 GB/s	153 GB/s
For models, at most	about 24 GB	about 24 GB
Models that run well	27 of 73	28 of 73
MLX seat class	Core	Core
Found in	MacBook Pro (14-inch, M4, 2024), iMac (24-inch, 2024, Two ports), iMac (24-inch, 2024, Four ports), Mac mini (2024), MacBook Air (13-inch, M4, 2025), MacBook Air (15-inch, M4, 2025)	MacBook Pro (14-inch, M5), MacBook Air (13-inch, M5), MacBook Air (15-inch, M5)

Speed with popular models

Tokens a second for one conversation, each chip at its largest memory size (32 GB and 32 GB), with the best version of the model that fits.

Model	M4	M5
Llama 3.1 8B Instruct8B	12–21 4-bit	11–19 6-bit
gpt-oss-20b21B · 3.6B active	28–51 native 4.25-bit	36–64 native 4.25-bit
Qwen3.6 35B A3B35B · 3B active	32–56 4-bit	41–72 4-bit
Qwen3.8 27B27B	2–4 6-bit	3–6 6-bit
Llama 3.3 70B Instruct70B	2–4 2-bit lower quality	3–6 2-bit lower quality
gpt-oss-120b117B · 5.1B active	too big	too big
Qwen3.5 122B A10B122B · 10B active	too big	too big
DeepSeek V4 Flash 0731284B · 13B active	too big	too big

Questions

Is the M5 faster than the M4 for LLMs?: The M5 reads memory about 27% faster than the M4 (153 vs 120 GB/s), so it writes about 27% faster with any model that fits both. Reading a long prompt depends on GPU compute instead, which this comparison doesn’t cover.
Which can run bigger models, the M4 or the M5?: Neither: both go up to 32 GB. At that size the M4 runs 27 and the M5 28 of 73 open models well; the faster one clears more of them at 10 tokens a second.

More comparisons

The M4 for local AI · The M5 for local AI · Compare every model · The best Mac for local LLMs

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.