readygpu

Beelink GTR9 Pro Ryzen AI Max+ 395 128GB vs Framework Desktop Ryzen AI Max+ 395 64GB

For running models locally, priced in Germany. At 32k context the first runs 72 models fully in memory, the second 59.

In short

The Beelink GTR9 Pro Ryzen AI Max+ 395 128GB has more memory for models: 128 GB against 64 GB. At 32k context the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB runs 72 of the models we track fully in memory and the Framework Desktop Ryzen AI Max+ 395 64GB 59; only the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB holds DeepSeek V4 Flash. On the models both run fully in memory, the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB and the Framework Desktop Ryzen AI Max+ 395 64GB are about as fast as each other (estimated).

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

Beelink GTR9 Pro Ryzen AI Max+ 395 128GB

Not priced here

Add to build
Framework Desktop Ryzen AI Max+ 395 64GB

Not priced here

Add to build

Specifications

SpecBeelink GTR9 Pro Ryzen AI Max+ 395 128GBFramework Desktop Ryzen AI Max+ 395 64GB
ChipRyzen AI Max+ 395 (Radeon 8060S)Ryzen AI Max+ 395 (Radeon 8060S)
Memory bus256-bit256-bit
Memory bandwidth256 GB/s256 GB/s
FP16 compute59.4 TFLOPS59.4 TFLOPS
SoftwareROCmROCm
Unified memory128 GB64 GB
GPU can use by default96 GB48 GB
GPU can use, tuned120 GB56 GB
Operating systemwindowslinux
ProcessorRyzen AI Max+ 395Ryzen AI Max+ 395
Storage2000 GB1000 GB

What each runs

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

ModelBeelink GTR9 Pro Ryzen AI Max+ 395 128GBFramework Desktop Ryzen AI Max+ 395 64GB
DeepSeek V4 Flash9tok/sUD-Q2_K_XLwon't run
Qwen 3.8 Flash Next3tok/sUD-Q3_K_XLwon't run
MiniMax M2.77tok/sUD-Q2_K_XLwon't run
Qwen 3.8 27B5tok/sQ8_05tok/sQ8_0
Qwen 3.5 122B-A10B9tok/sQ6_K20tok/sUD-Q2_K_XL
Qwen 3.6 27B5tok/sQ8_05tok/sQ8_0
gpt-oss 120B20tok/sMXFP4won't run
Qwen3 Coder Next17tok/sQ8_026tok/sQ4_K_M
Nemotron 3 Super 120B-A12B9tok/sQ5_K_Mwon't run
Gemma 4 31B4tok/sQ8_04tok/sQ8_0
Mistral Small 4 119B13tok/sUD-Q6_K27tok/sUD-Q2_K_XL
Qwen 3.6 35B-A3B22tok/sQ8_022tok/sQ8_0
Qwen 3.5 27B5tok/sQ8_05tok/sQ8_0
Gemma 4 26B-A4B17tok/sQ8_017tok/sQ8_0
GLM 4.5 Air4tok/sQ5_K_Mwon't run
Qwen 3.5 35B-A3B22tok/sQ8_022tok/sQ8_0
Qwen3 Next 80B-A3B21tok/sQ8_032tok/sQ4_K_M
GLM 4.7 Flash12tok/sQ8_012tok/sQ8_0
Devstral Small 2 24B5tok/sQ8_05tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B24tok/sUD-Q6_K_XL24tok/sUD-Q6_K_XL
gpt-oss 20B26tok/sMXFP426tok/sMXFP4
Kimi Linear 48B-A3B24tok/sQ8_030tok/sQ6_K
Qwen3 Coder 30B-A3B13tok/sQ8_013tok/sQ8_0
Qwen3 32B3tok/sQ8_03tok/sQ8_0
Gemma 4 12B10tok/sQ8_010tok/sQ8_0
Qwen 3.5 9B13tok/sQ8_013tok/sQ8_0
Mistral Small 3.2 24B5tok/sQ8_05tok/sQ8_0
Gemma 3 27B4tok/sQ8_04tok/sQ8_0
DeepSeek R1 Distill Qwen 32B3tok/sQ8_03tok/sQ8_0
Llama 3.3 70B2tok/sQ8_0won't run
Ministral 3 14B7tok/sQ8_07tok/sQ8_0
Qwen3 14B7tok/sQ8_07tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B10tok/sQ8_010tok/sQ8_0
Qwen3 8B10tok/sQ8_010tok/sQ8_0
Phi-4 14B6tok/sQ8_06tok/sQ8_0
Qwen 3.5 4B25tok/sQ8_025tok/sQ8_0
Ministral 3 8B10tok/sQ8_010tok/sQ8_0
Granite 4.2 8B10tok/sQ8_010tok/sQ8_0
Gemma 4 E4B15tok/sQ8_015tok/sQ8_0
LFM2.5 8B-A1B32tok/sQ8_032tok/sQ8_0
Llama 3.1 8B13tok/sQ6_K13tok/sQ6_K
Qwen2.5 0.5B151tok/sQ8_0151tok/sQ8_0
Qwen3.5 0.8B114tok/sQ8_0114tok/sQ8_0
Qwen3.5 2B76tok/sQ5_K_M76tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B50tok/sQ8_050tok/sQ8_0
MiniCPM5 2B34tok/sQ8_034tok/sQ8_0
LFM2.5 2.6B29tok/sQ8_029tok/sQ8_0
LFM2.5 VL 3B29tok/sQ8_029tok/sQ8_0
Mellum2.1 12B A2.5B Thinking27tok/sQ8_027tok/sQ8_0
granite vision 4.1 4b22tok/sQ8_022tok/sQ8_0
granite 4.2 3b21tok/sQ8_021tok/sQ8_0
Kimi VL A3B Thinking 250621tok/sQ8_021tok/sQ8_0
Kimi VL A3B20tok/smlx-8bit20tok/smlx-8bit
LFM2.5 24B A2B18tok/sQ8_018tok/sQ8_0
GLM 4 9B 041415tok/smlx-5bit15tok/smlx-5bit
DeepSeek R1 Distill Qwen 7B14tok/sQ8_014tok/sQ8_0
GLM 4.6V Flash12tok/sQ8_012tok/sQ8_0
DeepSeek R1 Distill Llama 8B11tok/sQ8_011tok/sQ8_0
GLM 4.1V 9B Thinking11tok/smlx-8bit11tok/smlx-8bit
Ministral 3 8B Reasoning 251210tok/sQ8_010tok/sQ8_0
DeepSeek R1 Distill Qwen 14B6tok/sQ8_06tok/sQ8_0
GLM 4.5V6tok/sQ4_K_Mwon't run
GLM 4.6V5tok/sQ4_K_Mwon't run
Devstral Small 25075tok/sQ8_05tok/sQ8_0
Magistral Small 25065tok/sQ8_05tok/sQ8_0
Magistral Small 25075tok/sQ8_05tok/sQ8_0
Magistral Small 25095tok/sQ8_05tok/sQ8_0
granite 4.2 30b4tok/sQ8_04tok/sQ8_0
DeepSeek R1 Distill Llama 70B2tok/sQ8_0won't run
Kimi Dev 72B2tok/sQ8_0won't run
Devstral 2 123B Instruct 25121tok/sQ5_K_Mwon't run
Mistral Medium 3.5 128B1tok/sQ5_K_Mwon'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 Beelink GTR9 Pro Ryzen AI Max+ 395 128GB or the Framework Desktop Ryzen AI Max+ 395 64GB?
Choose the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB to run bigger models.
Which can run bigger models, the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB or the Framework Desktop Ryzen AI Max+ 395 64GB?
At 32k context the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB runs 72 of the models we track fully in memory and the Framework Desktop Ryzen AI Max+ 395 64GB 59; only the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB holds DeepSeek V4 Flash.
Which is faster for local AI, the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB or the Framework Desktop Ryzen AI Max+ 395 64GB?
On the models both run fully in memory, the Beelink GTR9 Pro Ryzen AI Max+ 395 128GB and the Framework Desktop Ryzen AI Max+ 395 64GB are about as fast as each other (estimated).

More comparisons