Is the M5 Ultra faster than the M3 Ultra for LLMs?

The M5 Ultra reads memory about 47% faster than the M3 Ultra (1,200 vs 819 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 M3 Ultra or the M5 Ultra?

Neither: both go up to 512 GB. At that size the M3 Ultra runs 60 and the M5 Ultra 61 of 73 open models well; the faster one clears more of them at 10 tokens a second.

M3 Ultra vs M5 Ultra for local LLMs: memory, speed, which to buy

Side by side

	M3 Ultra	M5 Ultra
Memory sizes	96, 256, 512 GB	96, 256, 512 GB
Memory bandwidth	819 GB/s	1,200 GB/s
For models, at most	about 384 GB	about 384 GB
Models that run well	60 of 73	61 of 73
MLX seat class	Ultra	Ultra Latest
Found in	Mac Studio (2025)	Mac Studio (M5 Max or M5 Ultra)

Speed with popular models

Tokens a second for one conversation, each chip at its largest memory size (512 GB and 512 GB), with the best version of the model that fits.

Model	M3 Ultra	M5 Ultra
Llama 3.1 8B Instruct8B	43–77 8-bit	64–113 8-bit
gpt-oss-20b21B · 3.6B active	194–345 native 4.25-bit	284–505 native 4.25-bit
Qwen3.6 35B A3B35B · 3B active	115–205 8-bit	169–300 8-bit
Qwen3.8 27B27B	13–23 8-bit	19–33 8-bit
Llama 3.3 70B Instruct70B	5–9 8-bit	14–24 4-bit
gpt-oss-120b117B · 5.1B active	137–243 native 4.25-bit	200–356 native 4.25-bit
Qwen3.5 122B A10B122B · 10B active	35–62 8-bit	51–91 8-bit
DeepSeek V4 Flash 0731284B · 13B active	27–47 8-bit	39–70 8-bit

Questions

Is the M5 Ultra faster than the M3 Ultra for LLMs?: The M5 Ultra reads memory about 47% faster than the M3 Ultra (1,200 vs 819 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 M3 Ultra or the M5 Ultra?: Neither: both go up to 512 GB. At that size the M3 Ultra runs 60 and the M5 Ultra 61 of 73 open models well; the faster one clears more of them at 10 tokens a second.

More comparisons

The M3 Ultra for local AI · The M5 Ultra 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.