Is the M2 faster than the M1 for LLMs?

The M2 reads memory about 47% faster than the M1 (100 vs 68 GB/s), so it writes about 47% 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 M1 or the M2?

The M2: The M2 goes up to 24 GB of memory, the M1 to 16 GB. At its largest it runs 20 of 73 open models well, against 14.

M1 vs M2 for local LLMs: memory, speed, which to buy

Side by side

	M1	M2
Memory sizes	8, 16 GB	8, 16, 24 GB
Memory bandwidth	68 GB/s	100 GB/s
For models, at most	about 12 GB	about 18 GB
Models that run well	14 of 73	20 of 73
MLX seat class	Core	Core
Found in	MacBook Air (M1, 2020), MacBook Pro (13-inch, M1, 2020), Mac mini (M1, 2020), iMac (24-inch, M1, 2021)	MacBook Air (M2, 2022), MacBook Pro (13-inch, M2, 2022), MacBook Air (15-inch, M2, 2023), Mac mini (2023)

Speed with popular models

Tokens a second for one conversation, each chip at its largest memory size (16 GB and 24 GB), with the best version of the model that fits.

Model	M1	M2
Llama 3.1 8B Instruct8B	4–6 8-bit	10–18 4-bit
gpt-oss-20b21B · 3.6B active	19–34 3-bit lower quality	24–42 native 4.25-bit
Qwen3.6 35B A3B35B · 3B active	34–60 2-bit lower quality	35–62 3-bit lower quality
Qwen3.8 27B27B	4–6 2-bit lower quality	3–5 4-bit
Llama 3.3 70B Instruct70B	too big	too big
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 M2 faster than the M1 for LLMs?: The M2 reads memory about 47% faster than the M1 (100 vs 68 GB/s), so it writes about 47% 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 M1 or the M2?: The M2: The M2 goes up to 24 GB of memory, the M1 to 16 GB. At its largest it runs 20 of 73 open models well, against 14.

More comparisons

The M1 for local AI · The M2 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.