readygpu

Framework Desktop Ryzen AI Max+ 395 128GB vs Mac Studio M3 Ultra 96GB

For running models locally, priced in the US. At 64k context the first runs 71 models fully in memory, the second 65.

In short

The Framework Desktop Ryzen AI Max+ 395 128GB has more memory for models: 128 GB against 96 GB. At 64k context the Framework Desktop Ryzen AI Max+ 395 128GB runs 71 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 65; only the Framework Desktop Ryzen AI Max+ 395 128GB holds Qwen 3.8 Flash Next. On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.8× as fast as the Framework Desktop Ryzen AI Max+ 395 128GB (estimated, median across models). In the US the Framework Desktop Ryzen AI Max+ 395 128GB costs $3,449 at eBay (used) and the Mac Studio M3 Ultra 96GB $4,399 at eBay: the Framework Desktop Ryzen AI Max+ 395 128GB is $950 cheaper.

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

Framework Desktop Ryzen AI Max+ 395 128GB

$3,449

eBay, used (asking)

Add to build
Mac Studio M3 Ultra 96GB

$4,399

eBay

Add to build

Specifications

SpecFramework Desktop Ryzen AI Max+ 395 128GBMac Studio M3 Ultra 96GB
ChipRyzen AI Max+ 395 (Radeon 8060S)Apple M3 Ultra (80-core GPU)
Memory bus256-bit1024-bit
Memory bandwidth256 GB/s819 GB/s
FP16 compute59.4 TFLOPS28 TFLOPS
SoftwareROCmMetal
Unified memory128 GB96 GB
GPU can use by default96 GB72 GB
GPU can use, tuned120 GB88 GB
Operating systemlinuxmacos
ProcessorRyzen AI Max+ 395M3 Ultra 32-core
Storage1000 GB1000 GB

What each runs

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

ModelFramework Desktop Ryzen AI Max+ 395 128GBMac Studio M3 Ultra 96GB
Qwen 3.8 Flash Next3tok/sUD-Q3_K_XLwon't run
MiniMax M2.74tok/sUD-Q2_K_XLwon't run
Qwen 3.8 27B4tok/sQ8_016tok/sQ8_0
Qwen 3.5 122B-A10B9tok/sQ5_K_M51tok/sQ3_K_M
Qwen 3.6 27B4tok/sQ8_016tok/sQ8_0
gpt-oss 120B15tok/sMXFP458tok/sMXFP4
Qwen3 Coder Next15tok/sQ8_066tok/sQ6_K
Nemotron 3 Super 120B-A12B8tok/sQ5_K_M48tok/sUD-Q3_K_M
Gemma 4 31B4tok/sQ8_014tok/sQ8_0
Mistral Small 4 119B11tok/sUD-Q6_K54tok/sUD-Q4_K_M
Qwen 3.6 35B-A3B19tok/sQ8_071tok/sQ8_0
Qwen 3.5 27B4tok/sQ8_016tok/sQ8_0
Gemma 4 26B-A4B15tok/sQ8_056tok/sQ8_0
GLM 4.5 Air3tok/sQ5_K_M15tok/sQ3_K_M
Qwen 3.5 35B-A3B19tok/sQ8_071tok/sQ8_0
Qwen3 Next 80B-A3B18tok/sQ8_078tok/sQ6_K
GLM 4.7 Flash10tok/sQ8_036tok/sQ8_0
Devstral Small 2 24B4tok/sQ8_015tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B22tok/sUD-Q6_K_XL85tok/sUD-Q6_K_XL
Kimi Linear 48B-A3B23tok/sQ8_085tok/sQ8_0
gpt-oss 20B21tok/sMXFP479tok/sMXFP4
Qwen3 Coder 30B-A3B9tok/sQ8_033tok/sQ8_0
Qwen3 32B3tok/sQ8_010tok/sQ8_0
Gemma 4 12B10tok/sQ8_038tok/sQ8_0
Qwen 3.5 9B12tok/sQ8_045tok/sQ8_0
Gemma 3 27B4tok/sQ8_015tok/sQ8_0
Mistral Small 3.2 24B4tok/sQ8_015tok/sQ8_0
DeepSeek R1 Distill Qwen 32B3tok/sQ8_010tok/sQ8_0
Llama 3.3 70B2tok/sQ8_08tok/sQ5_K_M
Ministral 3 14B6tok/sQ8_021tok/sQ8_0
Qwen3 14B5tok/sQ8_020tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B8tok/sQ8_029tok/sQ8_0
Qwen3 8B8tok/sQ8_029tok/sQ8_0
Phi-4 14B5tok/sQ8_018tok/sQ8_0
Qwen 3.5 4B21tok/sQ8_080tok/sQ8_0
Ministral 3 8B8tok/sQ8_029tok/sQ8_0
Granite 4.2 8B7tok/sQ8_027tok/sQ8_0
Gemma 4 E4B14tok/sQ8_053tok/sQ8_0
LFM2.5 8B-A1B20tok/sQ8_075tok/sQ8_0
Llama 3.1 8B9tok/sQ6_K35tok/sQ6_K
Qwen2.5 0.5B105tok/sQ8_0398tok/sQ8_0
Qwen3.5 0.8B86tok/sQ8_0325tok/sQ8_0
Qwen3.5 2B62tok/sQ5_K_M236tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B37tok/sQ8_0141tok/sQ8_0
MiniCPM5 2B26tok/sQ8_097tok/sQ8_0
Mellum2.1 12B A2.5B Thinking23tok/sQ8_088tok/sQ8_0
LFM2.5 2.6B20tok/sQ8_077tok/sQ8_0
LFM2.5 VL 3B20tok/sQ8_077tok/sQ8_0
Kimi VL A3B Thinking 250616tok/sQ8_062tok/sQ8_0
Kimi VL A3B16tok/smlx-8bit60tok/smlx-8bit
granite vision 4.1 4b16tok/sQ8_059tok/sQ8_0
granite 4.2 3b15tok/sQ8_058tok/sQ8_0
GLM 4 9B 041413tok/smlx-5bit49tok/smlx-5bit
DeepSeek R1 Distill Qwen 7B12tok/sQ8_045tok/sQ8_0
LFM2.5 24B A2B11tok/sQ8_043tok/sQ8_0
GLM 4.6V Flash11tok/sQ8_042tok/sQ8_0
GLM 4.1V 9B Thinking10tok/smlx-8bit37tok/smlx-8bit
DeepSeek R1 Distill Llama 8B8tok/sQ8_031tok/sQ8_0
Ministral 3 8B Reasoning 25128tok/sQ8_029tok/sQ8_0
DeepSeek R1 Distill Qwen 14B5tok/sQ8_019tok/sQ8_0
GLM 4.5V4tok/sQ4_K_M15tok/sQ4_K_M
Devstral Small 25074tok/sQ8_015tok/sQ8_0
GLM 4.6V4tok/sQ4_K_Mwon't run
Magistral Small 25064tok/sQ8_015tok/sQ8_0
Magistral Small 25074tok/sQ8_015tok/sQ8_0
Magistral Small 25094tok/sQ8_015tok/sQ8_0
granite 4.2 30b3tok/sQ8_011tok/sQ8_0
DeepSeek R1 Distill Llama 70B2tok/sQ8_08tok/sQ5_K_M
Devstral 2 123B Instruct 25121tok/sQ4_K_Mwon't run
Kimi Dev 72B1tok/sQ8_0won't run
Mistral Medium 3.5 128B1tok/sQ4_K_Mwon't run
DeepSeek V4 Flashwon't runwon't run
GLM 5.3 Flashwon't runwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Zerowon't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.1 Terminuswon'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
GLM 4.6won'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 Framework Desktop Ryzen AI Max+ 395 128GB or the Mac Studio M3 Ultra 96GB?
Choose the Framework Desktop Ryzen AI Max+ 395 128GB to run bigger models. Choose the Mac Studio M3 Ultra 96GB for speed. Choose the Framework Desktop Ryzen AI Max+ 395 128GB to spend less.
Which can run bigger models, the Framework Desktop Ryzen AI Max+ 395 128GB or the Mac Studio M3 Ultra 96GB?
At 64k context the Framework Desktop Ryzen AI Max+ 395 128GB runs 71 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 65; only the Framework Desktop Ryzen AI Max+ 395 128GB holds Qwen 3.8 Flash Next.
Which is faster for local AI, the Framework Desktop Ryzen AI Max+ 395 128GB or the Mac Studio M3 Ultra 96GB?
On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.8× as fast as the Framework Desktop Ryzen AI Max+ 395 128GB (estimated, median across models).
What is the price difference between the Framework Desktop Ryzen AI Max+ 395 128GB and the Mac Studio M3 Ultra 96GB?
In the US the Framework Desktop Ryzen AI Max+ 395 128GB costs $3,449 at eBay (used) and the Mac Studio M3 Ultra 96GB $4,399 at eBay: the Framework Desktop Ryzen AI Max+ 395 128GB is $950 cheaper.

More comparisons