Hardware · Clusters · Example cluster

Two Ryzen AI Max+ 395 mini PCs (128 GB): what it runs together

Two 128 GB Ryzen AI Max+ 395 machines (Framework Desktop and similar) joined by a USB4 cable.

The PC answer to a pair of Mac Studios: about 190 GB together runs 200B to 360B mixture-of-experts models such as GLM 4.7 that neither machine holds alone, with llama.cpp RPC. Typing stays modest (about 12 tokens per second for GLM 4.7), because each machine reads memory at about 256 GB/s.

Change the devices How clustering works

192 GBusable memory together (the largest device has 96.0 GB)

25models only the cluster can run

27run better or faster than on any one device

429B (23.0B active)biggest model it adds: MiniMax M3.1 Flash (428B-A23B)

Connected by Thunderbolt 4 or USB4 cable. The wait for the first word is worked out for a 1,000-token prompt (a short chat).

The devices

Tool to use

llama.cpp RPC best estimate for 47 of the 52 models the cluster adds

Mixed devices you already own, a Mac, a gaming PC, a mini PC or a Jetson, when one GGUF model won't fit on any of them alone.

ggml-rpc-server --host 0.0.0.0 --port 50052 -c   # on each worker
llama-server -hf <model-GGUF> -ngl 99 --rpc 192.168.1.10:50052,192.168.1.11:50052   # on the main machine

Also works. The number is how many of the added models each one runs:

572 LLMs this cluster can run

Models the cluster adds come first, best match first, then the ones a single device already runs as well. Filter vs one device to pick. Speeds are estimates for one person chatting.

Quality Model Quant Memory Speed Runs Devices Tool vs one device
81Great quality (A) · 81 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. MiMo V2.6 Flash (310B-A15B)Xiaomi · 311B (15.0B active) Q3_K_M 169 GB 18.0 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek V4 Flash (April 2026)DeepSeek · 291B (13.0B active) Q3_K_M 159 GB 20.1 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek V4 Flash (284B-A13B)DeepSeek · 304B (13.0B active) Q3_K_M 166 GB 20.1 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
81Great quality (A) · 81 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. GLM-5.3 Flash (320B-A18B)Zhipu AI · 321B (18.0B active) Q3_K_M 174 GB 15.6 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. MiMo V2.5 (310B-A15B)Xiaomi · 311B (15.0B active) Q3_K_M 169 GB 18.0 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
79Great quality (A) · 79 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. Hunyuan Hy3 (295B-A21B)Tencent · 299B (21.0B active) Q3_K_M 166 GB 13.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
76Great quality (A) · 76 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. MiMo V2 Flash (309B-A15B)Xiaomi · 310B (15.0B active) Q3_K_M 169 GB 18.0 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
79Great quality (A) · 79 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. Naive N0.5 Flash (309B-A15.5B)Naive AI · 309B (15.5B active) Q3_K_M 168 GB 9.3 tok/s Slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split EXO Needs the clusterWon’t run on any one of them
66Good quality (B) · 66 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. Intern-S2 (397B-A17B)Shanghai AI Lab · 403B (17.0B active) Q2_K 178 GB 19.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
66Good quality (B) · 66 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. Nex N2 Pro (397B-A17B)Nex-AGI · 397B (17.0B active) Q2_K 176 GB 19.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
66Good quality (B) · 66 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.5 397B-A17BAlibaba · 403B (17.0B active) Q2_K 178 GB 19.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
65Good quality (B) · 65 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. Ornith 1.5 397B-A17BDeepReinforce · 403B (17.0B active) Q2_K 178 GB 19.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
67Good quality (B) · 67 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M3.1 Flash (428B-A23B)MiniMax · 429B (23.0B active) Q2_K 189 GB 15.6 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
66Good quality (B) · 66 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. K2-Horizon (375B-A23B)Institute of Foundation Models · 379B (23.0B active) Q2_K 170 GB 15.6 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
66Good quality (B) · 66 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M3 (427B)MiniMax · 427B (25.0B active) Q2_K 188 GB 14.6 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
65Good quality (B) · 65 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. GLM-4.7 (355B-A32B)Zhipu AI · 358B (32.0B active) Q2_K 163 GB 11.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. Trinity Large Thinking (400B-A13B)Arcee AI · 399B (13.0B active) Q2_K 177 GB 24.1 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
58Good quality (B) · 58 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking. GLM-4.6 (355B-A32B)Zhipu AI · 357B (32.0B active) Q2_K 162 GB 11.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
57Good quality (B) · 57 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking. GLM-4.5 (355B-A32B)Zhipu AI · 358B (32.0B active) Q2_K 163 GB 11.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
67Good quality (B) · 67 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking. ERNIE 4.5 300B-A47BBaidu · 300B (47.0B active) Q3_K_M 166 GB 3.3 tok/s Slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split EXO Needs the clusterWon’t run on any one of them
50Fair quality (C) · 50 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. Llama 4 Maverick (17B x 128E)Meta · 402B (17.0B active) Q2_K 179 GB 19.8 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. Hermes 4 405BNous Research · 406B Q2_K 183 GB 1.1 tok/s Very slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
53Fair quality (C) · 53 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 65/100 (quality, fit, speed and memory use) sets the ranking. Tülu 3 405BAi2 · 406B Q2_K 183 GB 1.1 tok/s Very slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
50Fair quality (C) · 50 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking. Llama 3.1 405B InstructMeta · 406B Q2_K 183 GB 1.1 tok/s Very slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
42Fair quality (C) · 42 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking. Grok-1 (314B)xAI · 314B (86.0B active) Q3_K_M 173 GB 3.9 tok/s Slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Needs the clusterWon’t run on any one of them
93Excellent quality (A+) · 93 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 93/100 (quality, fit, speed and memory use) sets the ranking. Qwen3.8 Flash Next (125B-A6B)Alibaba · 180B (6.0B active) Q5_K_M 145 GB 24.7 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs great on Ryzen AI 1, 60.1 tok/s
88Excellent quality (A+) · 88 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M2.7 (230B-A10B)MiniMax · 229B (10.0B active) Q4_K_M 158 GB 20.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 33.2 tok/s
90Excellent quality (A+) · 90 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. Step 3.7 Flash (198B-A11B)StepFun · 201B (11.0B active) Q5_K_M 160 GB 15.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 35.9 tok/s
87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M2.5 (230B-A10B)MiniMax · 229B (10.0B active) Q4_K_M 158 GB 20.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 33.2 tok/s
89Excellent quality (A+) · 89 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. Step 3.5 Flash (196B-A11B)StepFun · 199B (11.0B active) Q5_K_M 159 GB 15.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 35.9 tok/s
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. Nemotron 3.5 Super VL 120B-A12BNVIDIA · 124B (12.0B active) Q6_K 119 GB 12.8 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 21.0 tok/s
86Excellent quality (A+) · 86 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M2.1 (230B-A10B)MiniMax · 229B (10.0B active) Q4_K_M 158 GB 20.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 33.2 tok/s
89Excellent quality (A+) · 89 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. Inkling Small (276B-A12B)Thinking Machines Lab · 266B (12.0B active) Q4_K_M 179 GB 17.4 tok/s Runs welltight fit Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Slow on Ryzen AI 1, 28.2 tok/s
84Great quality (A) · 84 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. MiniMax M2 (230B-A10B)MiniMax · 229B (10.0B active) Q4_K_M 158 GB 20.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 33.2 tok/s
87Excellent quality (A+) · 87 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Nemotron 3 Super 120B-A12BNVIDIA · 124B (12.0B active) Q6_K 118 GB 12.8 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 21.0 tok/s
86Excellent quality (A+) · 86 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Qwen3 235B-A22B Thinking 2507Alibaba · 235B (22.0B active) Q4_K_M 162 GB 10.6 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.2 tok/s
85Excellent quality (A+) · 85 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking. Qwen3 235B-A22B Instruct 2507Alibaba · 235B (22.0B active) Q4_K_M 162 GB 10.6 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.2 tok/s
84Great quality (A) · 84 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3-VL 235B-A22B InstructAlibaba · 236B (22.0B active) Q4_K_M 162 GB 10.6 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.2 tok/s
83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. Qwen3 235B-A22BAlibaba · 235B (22.0B active) Q4_K_M 161 GB 10.6 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.2 tok/s
83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. K-EXAONE 236B-A23BLG AI Research · 237B (23.0B active) Q4_K_M 162 GB 10.2 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 15.5 tok/s
85Excellent quality (A+) · 85 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. Solar Open 2 250BUpstage · 250B (15.0B active) Q4_K_M 171 GB 7.6 tok/s Slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split EXO Better on the clusterAlone: Slow on Ryzen AI 1, 23.0 tok/s
79Great quality (A) · 79 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. Baichuan M3 235B-A22BBaichuan · 235B (22.0B active) Q4_K_M 162 GB 10.6 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.2 tok/s
75Great quality (A) · 75 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. dots.llm1 Instructrednote · 143B (14.0B active) Q6_K 142 GB 11.3 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 23.1 tok/s
73Great quality (A) · 73 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. Command A+ (218B-A25B)Cohere · 219B (25.0B active) Q3_K_M 124 GB 11.9 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 14.3 tok/s
72Great quality (A) · 72 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking. InternVL3.5 241B-A28BShanghai AI Lab · 241B (28.0B active) Q3_K_M 136 GB 10.8 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Slow on Ryzen AI 1, 12.9 tok/s
70Great quality (A) · 70 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek Coder V2 (236B)DeepSeek · 236B (21.0B active) Q4_K_M 161 GB 11.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.9 tok/s
68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek Coder V2 Instruct (0614)DeepSeek · 236B (21.0B active) Q4_K_M 161 GB 11.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.9 tok/s
68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek V2.5 1210DeepSeek · 236B (21.0B active) Q4_K_M 161 GB 11.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.9 tok/s
66Good quality (B) · 66 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. DeepSeek V2.5DeepSeek · 236B (21.0B active) Q4_K_M 161 GB 11.0 tok/s Runs well Ryzen AI 1 + Ryzen AI 2Pipeline split llama.cpp RPC Better on the clusterAlone: Runs well on Ryzen AI 1, 16.9 tok/s
81Great quality (A) · 81 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking. Llama 3.1 Nemotron Ultra 253B v1NVIDIA · 253B Q4_K_M 175 GB 0.5 tok/s Very slowtight fit Ryzen AI 1 + Ryzen AI 2Pipeline split EXO Better on the clusterAlone: Very slow on Ryzen AI 1, 1.5 tok/s

Other example clusters

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.

Was this page helpful?