what/llm/can/i/run
gpus/apple/m3-ultra-96gb

M3 Ultra 96GB

Released 2025-03-12 · Street $3,999
vram
96 GB
unified mem
bandwidth
800
GB/s
models fit
18
top · FP16
power
120 W
low

Spec sheet · top pick

Hardware spec
M3 Ultra 96GB
Apple · Apple Silicon · M3
vram
96 GB unified
bandwidth
800 GB/s
tdp
120 W
released
2025-03-12
msrp
$3,999
street
~$3,999
★ Sweet spot for this card
gemma 3 27b
FP16 · GGUF · 57.2 GB
bench50.6 · conf A
vram57.2 / 96 GB
speed14 tok/s
Best benchmark per GB of VRAM in your tier. FP16wastes <3% quality vs FP16, leaves headroom for 32k context.
Full model inspect →

Top 10 models on this GPU

#modelquantvramtok/sbench ↓fit
01gemma 3 27bFP1657.2 GB1450.6Great
02Qwen2.5 Coder 32BFP1667.7 GB1248.8Great
03Gemma 2 27BFP1657.2 GB1446.2Great
04deepseek r1 distill llama 70bQ8_074 GB1151.7Great
05Qwen3 32BFP1667.7 GB1240.0Great
06Gemma 2 9BFP1619.4 GB4135.0Great
07Gemma 3 12BFP1625.7 GB3131.3Great
08deepseek r1 distill qwen 32bFP1667.7 GB1233.3Great
09Gemma 3 4BFP168.9 GB9015.1Great
10Mixtral 8x7B v0.1Q8_049.5 GB1613.3Great
Showing 10 of 18 models that fit · see full ranked list →

Upgrade options

★ Best value
+H200 SXM5
141 GB · ~$45,000 · +$41,001 over M3 Ultra 96GB
Unlocks +1 models not in current rank
See H200 SXM5
Bigger tier
+M1 Ultra 128GB
128 GB · ~$4,000 · +$1 over M3 Ultra 96GB
Unlocks +1 models not in current rank
See M1 Ultra 128GB
Premium tier
+M2 Ultra 128GB
128 GB · ~$5,200 · +$1,201 over M3 Ultra 96GB
Unlocks +1 models not in current rank
See M2 Ultra 128GB

Also compare

H100 PCIe
80 GB · NVIDIA · 600% more
H100 SXM5
80 GB · NVIDIA · 900% more
A100 80GB
80 GB · NVIDIA · 175% more
M2 Max 96GB
96 GB · Apple unified · 12% cheaper
Not running a M3 Ultra 96GB?
Tell us what you actually have — we'll re-rank everything for your machine.
Change my machine