readygpu

ASUS Ascent GX10 128GB vs Mac Studio M3 Ultra 96GB

For running models locally, priced in the UK. At 32k context the first runs 69 models fully in memory, the second 63.

In short

The ASUS Ascent GX10 128GB has more memory for models: 128 GB against 96 GB. At 32k context the ASUS Ascent GX10 128GB runs 69 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 63; only the ASUS Ascent GX10 128GB holds DeepSeek V4 Flash. On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.1× as fast as the ASUS Ascent GX10 128GB (estimated, median across models). In the UK the ASUS Ascent GX10 128GB costs £4,500 at Amazon and the Mac Studio M3 Ultra 96GB £5,499 at eBay: the ASUS Ascent GX10 128GB is £999 cheaper.

Updated . Speeds are estimates; how they are worked out.

ASUS Ascent GX10 128GB

£4,500

Amazon

Add to build
Mac Studio M3 Ultra 96GB

£5,499

eBay

Add to build

Specifications

SpecASUS Ascent GX10 128GBMac Studio M3 Ultra 96GB
ChipGB10 Grace BlackwellApple M3 Ultra (80-core GPU)
Memory bus256-bit1024-bit
Memory bandwidth273 GB/s819 GB/s
FP16 compute62 TFLOPS28 TFLOPS
SoftwareCUDAMetal
Unified memory128 GB96 GB
GPU can use by default119 GB72 GB
GPU can use, tuned122 GB88 GB
Operating systemdgx-osmacos
Processor20-core Arm (GB10)M3 Ultra 32-core
Storage1000 GB1000 GB

What each runs

Estimated decode speed with the most precise quant that fits each one est.

ModelASUS Ascent GX10 128GBMac Studio M3 Ultra 96GB
DeepSeek V4 Flash11tok/sUD-Q2_K_XLwon't run
GLM 5.3 Flash11tok/sUD-Q2_K_XLwon't run
Qwen 3.8 Flash Next2tok/sQ4_K_Mwon't run
MiniMax M2.78tok/sUD-Q3_K_Mwon't run
Qwen 3.8 27B5tok/sQ8_017tok/sQ8_0
Qwen 3.5 122B-A10B11tok/sQ6_K58tok/sQ3_K_M
Qwen 3.6 27B6tok/sQ8_017tok/sQ8_0
gpt-oss 120B23tok/sMXFP474tok/sMXFP4
Qwen3 Coder Next20tok/sQ8_078tok/sQ6_K
Nemotron 3 Super 120B-A12B9tok/sUD-Q6_K50tok/sUD-Q3_K_M
Gemma 4 31B5tok/sQ8_015tok/sQ8_0
Mistral Small 4 119B15tok/sUD-Q6_K62tok/sUD-Q4_K_M
Qwen 3.6 35B-A3B27tok/sQ8_084tok/sQ8_0
Qwen 3.5 27B6tok/sQ8_017tok/sQ8_0
Gemma 4 26B-A4B20tok/sQ8_063tok/sQ8_0
GLM 4.5 Air4tok/sQ8_021tok/sQ3_K_M
Qwen 3.5 35B-A3B27tok/sQ8_084tok/sQ8_0
Qwen3 Next 80B-A3B25tok/sQ8_097tok/sQ6_K
GLM 4.7 Flash14tok/sQ8_045tok/sQ8_0
Devstral Small 2 24B6tok/sQ8_018tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B28tok/sUD-Q6_K_XL89tok/sUD-Q6_K_XL
gpt-oss 20B31tok/sMXFP499tok/sMXFP4
Kimi Linear 48B-A3B29tok/sQ8_091tok/sQ8_0
Qwen3 Coder 30B-A3B16tok/sQ8_050tok/sQ8_0
Qwen3 32B4tok/sQ8_012tok/sQ8_0
Gemma 4 12B13tok/sQ8_039tok/sQ8_0
Qwen 3.5 9B16tok/sQ8_050tok/sQ8_0
Mistral Small 3.2 24B6tok/sQ8_018tok/sQ8_0
Gemma 3 27B5tok/sQ8_017tok/sQ8_0
DeepSeek R1 Distill Qwen 32B4tok/sQ8_012tok/sQ8_0
Llama 3.3 70B2tok/sQ8_09tok/sQ5_K_M
Ministral 3 14B9tok/sQ8_027tok/sQ8_0
Qwen3 14B8tok/sQ8_025tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B13tok/sQ8_039tok/sQ8_0
Qwen3 8B13tok/sQ8_039tok/sQ8_0
Phi-4 14B8tok/sQ8_024tok/sQ8_0
Qwen 3.5 4B30tok/sQ8_095tok/sQ8_0
Ministral 3 8B13tok/sQ8_039tok/sQ8_0
Granite 4.2 8B12tok/sQ8_036tok/sQ8_0
Gemma 4 E4B19tok/sQ8_058tok/sQ8_0
LFM2.5 8B-A1B38tok/sQ8_0119tok/sQ8_0
Llama 3.1 8B16tok/sQ6_K49tok/sQ6_K
Qwen2.5 0.5B181tok/sQ8_0570tok/sQ8_0
Qwen3.5 0.8B137tok/sQ8_0432tok/sQ8_0
Qwen3.5 2B91tok/sQ5_K_M287tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B60tok/sQ8_0188tok/sQ8_0
MiniCPM5 2B41tok/sQ8_0130tok/sQ8_0
LFM2.5 2.6B35tok/sQ8_0109tok/sQ8_0
LFM2.5 VL 3B35tok/sQ8_0109tok/sQ8_0
Mellum2.1 12B A2.5B Thinking32tok/sQ8_0101tok/sQ8_0
granite vision 4.1 4b27tok/sQ8_084tok/sQ8_0
granite 4.2 3b26tok/sQ8_081tok/sQ8_0
Kimi VL A3B24tok/smlx-8bit75tok/smlx-8bit
LFM2.5 24B A2B21tok/sQ8_066tok/sQ8_0
DeepSeek R1 Distill Qwen 7B17tok/sQ8_053tok/sQ8_0
GLM 4.6V Flash15tok/sQ8_047tok/sQ8_0
DeepSeek R1 Distill Llama 8B13tok/sQ8_042tok/sQ8_0
GLM 4.1V 9B Thinking13tok/smlx-8bit41tok/smlx-8bit
Ministral 3 8B Reasoning 251213tok/sQ8_039tok/sQ8_0
DeepSeek R1 Distill Qwen 14B8tok/sQ8_024tok/sQ8_0
Devstral Small 25076tok/sQ8_018tok/sQ8_0
Magistral Small 25066tok/sQ8_018tok/sQ8_0
Magistral Small 25076tok/sQ8_018tok/sQ8_0
Magistral Small 25096tok/sQ8_018tok/sQ8_0
GLM 4.5V5tok/sQ8_022tok/sQ4_K_M
GLM 4.6V5tok/sQ8_0won't run
granite 4.2 30b4tok/sQ8_013tok/sQ8_0
DeepSeek R1 Distill Llama 70B2tok/sQ8_09tok/sQ5_K_M
Mistral Medium 3.5 128B2tok/sQ6_Kwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.2won't runwon't run
DeepSeek V4 Flash 0731won't runwon't run
DeepSeek V4 Flash Vision Expwon't runwon't run
Devstral 2 123B Instruct 2512won't runwon't run
GLM 4.7won't runwon't run
GLM 5.1won't runwon't run
GLM 5.2won't runwon't run
GLM 5.3won't runwon't run
Kimi K2won't runwon't run
Kimi K2 Instruct 0905won't runwon't run
Kimi K2 Thinkingwon't runwon't run
Kimi K2.5won't runwon't run
MiniMax M2won't runwon't run
MiniMax M2.1won't runwon't run
MiniMax M2.5won't runwon't run
MiniMax M3won't runwon't run

Questions

Which is better for local AI, the ASUS Ascent GX10 128GB or the Mac Studio M3 Ultra 96GB?
Choose the ASUS Ascent GX10 128GB to run bigger models. Choose the Mac Studio M3 Ultra 96GB for speed. Choose the ASUS Ascent GX10 128GB to spend less.
Which can run bigger models, the ASUS Ascent GX10 128GB or the Mac Studio M3 Ultra 96GB?
At 32k context the ASUS Ascent GX10 128GB runs 69 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 63; only the ASUS Ascent GX10 128GB holds DeepSeek V4 Flash.
Which is faster for local AI, the ASUS Ascent GX10 128GB or the Mac Studio M3 Ultra 96GB?
On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.1× as fast as the ASUS Ascent GX10 128GB (estimated, median across models).
What is the price difference between the ASUS Ascent GX10 128GB and the Mac Studio M3 Ultra 96GB?
In the UK the ASUS Ascent GX10 128GB costs £4,500 at Amazon and the Mac Studio M3 Ultra 96GB £5,499 at eBay: the ASUS Ascent GX10 128GB is £999 cheaper.

More comparisons