RigBaseBuildsGPUsCan it run?GuidesBest rig for…By budgetRent before you buySubmit a buildNewsletterAdvertise

All 105 builds

Every home AI rig in RigBase, A–Z. Filter by budget and model on the homepage, or browse by model, by GPU or by VRAM size. Download all of it (CSV/JSON, CC BY 4.0).

(MTP, DFlash and similar; not comparable 1:1 with plain decode)

BuildGPU(s)Memory GBMeasured tok/s
$2K EPYC 64-core, 512 GB DDR4, CPU-onlyNone (CPU inference)0DeepSeek R1 671B: 4.31
2x DGX Spark running GLM-5.3 Flash NVFP42x Nvidia GB10 (unified, DGX Spark)256GLM-5.3 Flash: 33.0
GLM-5.3 Flash: 56.5
2x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open 816B2x Apple M5 Ultra (unified, Mac Studio)Muse Spark Open 816B-A42B: 24.99
Muse Spark Open 816B-A42B: 15.74
Mistral Medium 3.5 128B: 13.69
Qwen3.8 27B: 33.27
2x RTX 3090 (llm-speed signed run, Oct 2026): Llama 3.3 70B Q4_K_M at 19.1 tok/s2x RTX 309048Llama 3.3 70B: 19.1
4x DGX Spark cluster running GLM-5.2 (unpruned)4x Nvidia GB10 (unified, DGX Spark)512GLM-5.2: 32.5 ⚡
4x DGX Spark ring running GLM-5.3 Flash NVFP44x Nvidia GB10 (unified, DGX Spark)512GLM-5.3 Flash: 61.7
GLM-5.3 Flash: 105.3
4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s4x Apple M5 Ultra (unified, Mac Studio)Muse Spark Open 816B-A42B: 36.73
Mistral Medium 3.5 128B: 23.37
Qwen3.8 27B: 50.89
4x RTX 3090 + 32-core Threadripper, 192 GB DDR4 ⚠ unchecked4x RTX 309096Qwen3 235B-A22B: 12-14
4x RTX 3090 + EPYC 7443, 256 GB DDR4 8-channel ⚠ unchecked4x RTX 309096Qwen3 235B-A22B: 18
4x RTX 3090 homelab on Threadripper 2920X ⚠ unchecked4x RTX 3090 (~$625 each, used)96no speed data
4x RTX 3090 in a 4U rack case (AM5) ⚠ unchecked4x Dell OEM RTX 309096no speed data
4x RTX 3090 lab: Qwen3.6 on vLLM vs llama.cpp4x RTX 309096Qwen3.6 35B-A3B: 115-133
Qwen3.6 35B-A3B: 149
Qwen3.6 35B-A3B: 225-231 ⚡
8x RTX 3090 serving DeepSeek V4 Flash (LimeChain vLLM kit)8x RTX 3090 24 GB192DeepSeek V4 Flash: 95.52
AmpereOne A192-32X 512GB (CPU only)None (CPU only)DeepSeek R1 671B: 4.18
Llama 3.1 70B: 3.86
Llama 3.1 405B: 0.9
Llama 3.2 3B: 23.52
Apple M1 Max 64GBApple M1 Max (unified)64gpt-oss 20B: 75.15
Apple M4 Max 36GBApple M4 Max (unified)36gpt-oss 20B: 92.36
Dell Pro Max with GB10 (Nvidia DGX Spark)Nvidia GB10 (unified)gpt-oss 20B: 87.97
DeepSeek R1 Distill 14B: 23.92
Llama 3.1 70B: 4.71
Llama 3.2 3B: 97.93
DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTPNvidia GB10 (unified)128Qwen3.6 35B-A3B: 113 ⚡
Qwen3.6 35B-A3B: 86 ⚡
DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/sNvidia GB10 (unified)128Qwen3.6 35B-A3B: 102.3 ⚡
Qwen3.6 35B-A3B: 78.1
gpt-oss 20B: 96.2
Nemotron 3.5 Lightning 30B-A3B: 94.4 ⚡
Nemotron 3.5 Lightning 30B-A3B: 70.3
Nemotron 3 Nano Omni 30B: 58.8
Gemma 4 26B-A4B: 56.5 ⚡
Gemma 4 26B-A4B: 30.8
gpt-oss 120B: 44.4
Qwen3.8 Flash-Next: 26.8 ⚡
Nemotron 3 Super 120B-A12B: 23.0 ⚡
Nemotron 3 Super 120B-A12B: 16.1
Qwen3.8 27B: 19.0 ⚡
Qwen3.8 27B: 11.0
Llama 3.1 8B: 26.4
DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runsNvidia GB10 (unified)128Nemotron 3.5 Super VL 120B-A12B: 19.06
Nemotron 3.5 Super VL 120B-A12B: 38.07 ⚡
Nemotron 3.5 Bolt 74B-A9B: 39.47
Nemotron 3.5 Bolt 74B-A9B: 61.61 ⚡
DGX Spark with Qwen3.6 27B NVFP4 + DFlash speculative decodingNvidia GB10 (unified)128Qwen3.6 27B: 32.83 ⚡
Qwen3.6 27B: 7.85
Dual Intel Arc Pro B70 (neural.download lab)2x Intel Arc Pro B70 32 GB64Qwen3.8 Flash-Next: 58.56
Qwen3.8 27B: 46.01 ⚡
Dual RTX 3090 on Ryzen 3700X (X470) ⚠ unchecked2x Zotac RTX 309048Llama 3 70B: 16.95
Llama 3 70B: 21.2
Dual RTX 5070 Ti on Ryzen 9 9900X (capy-flash maintainer rig)2x RTX 5070 Ti 16 GB32Qwen3.8 Flash-Next: 182
Qwen3.8 Flash-Next: 155
Qwen3.8 27B: 139 ⚡
Qwen3.8 27B: 230 ⚡
Dual RTX 5080 on EPYC 7F52 (capy-flash user report)2x RTX 5080 16 GB32Qwen3.8 Flash-Next: 141.3
Qwen3.8 Flash-Next: 144.4
Dual RTX PRO 4000 Blackwell on Threadripper PRO 5975WX2x RTX PRO 4000 Blackwell 24 GB48Qwen3.8 Flash-Next: 100
Dual Tesla V100 32 GB on EPYC 7402 (bnk engine)2x Tesla V100 32 GB (PG500-216, sm_70)64Qwen3.8 27B: 42.4
Qwen3.8 27B: 104 ⚡
Qwen3.8 Flash-Next: 114
Qwen3.8 Flash-Next: 121
DeepSeek V4.1 Flash: 13
Dual used RTX 3090, ~2000 EUR total ⚠ unchecked2x RTX 3090 (used)48Llama 3.3 70B: 20
EPYC + single RTX 3090 (hybrid) ⚠ uncheckedRTX 309024DeepSeek R1 671B: 8.97
EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engineRTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card120Qwen3.8 Flash-Next: 102.889 ⚡
Qwen3.8 Flash-Next: 205.263 ⚡
Qwen3.8 Flash-Next: 76.337
Qwen3.8 Flash-Next: 22.657 ⚡
Qwen3.8 Flash-Next: 98.291 ⚡
EPYC 7F52 jank rig: 4x3090 + 4090 + 4080 (136 GB VRAM) ⚠ unchecked4x RTX 3090, RTX 4090, RTX 4080136Llama 3.1 70B: 9
Framework Desktop (AMD Strix Halo) mainboardAMD Ryzen AI Max+ 395 (Radeon 8060S iGPU; Ollama runs CPU only, llama.cpp runs on iGPU via Vulkan)DeepSeek R1 Distill 8B: 19.49
DeepSeek R1 Distill 70B: 2.43
Llama 3.1 70B: 4.97
Llama 3.2 3B: 88.14
HP OmniBook Ultra 16 (RTX Spark N1X, 128GB) laptop: local AI benchmarksNvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)128Qwen3.6 35B-A3B: 89.5
gpt-oss 20B: 58.0
Llama 3.1 8B: 56.5
HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)128Qwen3.6 35B-A3B: 59.4
gpt-oss 120B: 44.7
Qwen3.6 35B-A3B: 54.4
Gemma 4 26B-A4B: 51.8
Qwen3.6 27B: 11.4
Gemma 4 31B: 8.8
i7-12700K + RTX 5060 Ti 16GB + RTX 3060 12GB: Qwen3.8 27B from 21 to 51 tok/s (split, MTP, single-GPU)RTX 5060 Ti 16GB + RTX 3060 12GB28Qwen3.8 27B: 21.2
Qwen3.8 27B: 37.6 ⚡
Qwen3.8 27B: 28.0
Qwen3.8 27B: 51.4 ⚡
Intel Core Ultra 265K + Radeon AI Pro R9700AMD Radeon AI Pro R970032gpt-oss 20B: 163.26
DeepSeek R1 Distill 14B: 53.72
Llama 3.2 3B: 195.91
Intel Core Ultra 265K + RTX 2080 TiRTX 2080 Ti11DeepSeek R1 Distill 14B: 45.06
Llama 3.2 3B: 180.38
Intel Core Ultra 265K + RTX 3060RTX 306012gpt-oss 20B: 83.96
DeepSeek R1 Distill 14B: 29.77
Llama 3.2 3B: 122.85
Intel Core Ultra 265K + RTX 3080 TiRTX 3080 Ti12DeepSeek R1 Distill 14B: 70.74
Llama 3.2 3B: 263.24
Intel Core Ultra 265K + RTX 4070 TiRTX 4070 Ti12gpt-oss 20B: 163.14
DeepSeek R1 Distill 14B: 49.74
Llama 3.2 3B: 196.15
Intel Core Ultra 265K + RTX 4090RTX 409024gpt-oss 20B: 279.65
DeepSeek R1 Distill 14B: 92.49
Llama 3.2 3B: 334.67
Intel Core Ultra 265K + RTX A4000RTX A400016gpt-oss 20B: 93.55
Llama 3.2 3B: 149.44
Intel i9-13900K + RTX 5090RTX 509032Llama 3.2 3B: 271.4
M4 Mac mini (10-core) 32GBApple M4 (unified)32Llama 3.2 3B: 41.31
Mac Studio 64 GB ⚠ uncheckedApple unified memory64Llama 3.3 70B: 11
Mac Studio M1 UltraApple M1 UltraLlama 3.1 8B: 37.92
Mac Studio M1 Ultra (64-core GPU) 128GBApple M1 Ultra 64-core GPU (unified)128gpt-oss 20B: 100.43
DeepSeek R1 Distill 14B: 41.51
Llama 3.1 70B: 9.84
Llama 3.2 3B: 127.6
Mac Studio M2 Ultra 76-core GPU, 192 GBApple M2 Ultra 76-core GPU192Llama 2 7B: 125.21
gpt-oss 20B: 116.08
gpt-oss 120B: 79.68
Mac Studio M3 Ultra 512GB (Geerling test)Apple M3 Ultra (unified)512DeepSeek R1 671B: 19.89
gpt-oss 20B: 115.29
gpt-oss 20B: 115.52
Llama 3.1 70B: 14.08
Llama 3.1 405B: 1.96
DeepSeek R1 Distill 14B: 51.85
Llama 3.2 3B: 154.6
Mac Studio M3 Ultra 512GB: Muse Spark Open 816B (Hugging Face test)Apple M3 Ultra (unified)512Muse Spark Open 816B-A42B: 18
Muse Spark Open 816B-A42B: 21
Mac Studio M3 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark)Apple M3 Ultra (unified)512Qwen3.8 27B: 12.46
Mistral Medium 3.5 128B: 4.95
Mac Studio M3 Ultra, 512 GB ⚠ uncheckedApple M3 Ultra512DeepSeek R1 671B: 18.43
Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLXApple M5 Max (unified)36Qwen3.6 35B-A3B: 173.4 ⚡
Qwen3.6 35B-A3B: 145.8
Qwen3.6 35B-A3B: 124.5
Qwen3.8 27B: 37.9 ⚡
Qwen3.8 27B: 25.3
Qwen3.8 27B: 26.0
Qwen3.5 9B: 115.5 ⚡
Qwen3.5 9B: 78.0
Mac Studio M5 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark)Apple M5 Ultra (unified)512Qwen3.8 27B: 19.69
Mistral Medium 3.5 128B: 7.7
MacBook Pro M1 Max 32-core GPU, 64 GBApple M1 Max 32-core GPU64Llama 3 8B: 34.49
MacBook Pro M2 MaxApple M2 MaxLlama 3.1 8B: 28.92
MacBook Pro M3 Max 40-core GPU, 64 GBApple M3 Max 40-core GPU64Llama 3 8B: 50.74
MacBook Pro M4 Pro 48 GB: Flint 40B-A2.5B 4-bit at 127.9 tok/s (Kapitsa), 126.6 (rec), 79.6 (llama.cpp)Apple M4 Pro (20-core GPU, unified)48Flint 40B-A2.5B: 127.9
Flint 40B-A2.5B: 126.56
Flint 40B-A2.5B: 79.57
MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166Apple M5 Max (unified)128Qwen3.6 35B-A3B: 271 ⚡
Qwen3.6 35B-A3B: 240 ⚡
Qwen3.6 35B-A3B: 107
Gemma 4 26B-A4B: 166 ⚡
Qwen3.8 27B: 31
Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center)Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)Nemotron 3 Nano Omni 30B: 63.9
GLM 4.7 Flash: 55.6
Qwen3.5 122B: 35.2
Muse Glimmer 30B: 15.9
Qwen3.8 27B: 14.5
Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/sAMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)192gpt-oss 120B: 55.8
gpt-oss 120B: 45.1
MiniMax M2.7: 29.5
MiniMax M2.7: 17.3
DeepSeek V4 Flash: 17.4
DeepSeek V4 Flash: 12.2
Qwen3.8 27B: 8
Minisforum MS-S1 MAX (Strix Halo, 128 GB): Qwen3.6 35B-A3B at 62 tok/s on llama.cpp VulkanAMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)128Qwen3.6 35B-A3B: 62
Qwen3.6 35B-A3B: 54
Qwen3.6 35B-A3B: 51
Qwen3.6 35B-A3B: 93 ⚡
Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig)AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used)32Qwen3.6 35B-A3B: 102.39
Qwen3.6 35B-A3B: 90.52
Qwen3.6 35B-A3B: 66.43
Qwen3.6 35B-A3B: 44.4
Qwen3.6 35B-A3B: 109.73
Qwen3.6 35B-A3B: 117.69
Radeon AI Pro R9700: Qwen3.6 27B on Vulkan, HIP and Ollama, with and without MTPAMD Radeon AI Pro R9700 (1 of 2 cards in the rig used)32Qwen3.6 27B: 66.0 ⚡
Qwen3.6 27B: 31.9
Qwen3.6 27B: 27.1
Qwen3.6 27B: 26.7
Radeon AI Pro R9700: Qwen3.8 27B on llama.cpp, Ollama and vLLM (Vadi Taslim)AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used)32Qwen3.8 27B: 57 ⚡
Qwen3.8 27B: 30
Qwen3.8 27B: 22
Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTPAMD Radeon RX 7900 XTX24Qwen3.6 35B-A3B: 87.2
Qwen3.6 35B-A3B: 79.6
Qwen3.6 35B-A3B: 135.2 ⚡
Qwen3.6 35B-A3B: 66.0
Qwen3.6 35B-A3B: 108.2 ⚡
Qwen3.6 35B-A3B: 38.3
RTX 3060 12GB: 2026 models (llamabench community)RTX 306012Gemma 4 12B: 38.5
LFM2.5 24B-A2B: 125
Qwen3.8 27B: 5.2-6.7
RTX 3090 (llm-speed signed run, Oct 2026): Flint 40B-A2.5B Q3_K_XL at 128.9 tok/sRTX 309024Flint 40B-A2.5B: 128.9
RTX 3090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 207 tok/s, Qwen3.6 27B 43 tok/sRTX 309024gpt-oss 20B: 207.2
Qwen3.6 27B: 42.7
GLM 4.7 Flash: 134.6
RTX 3090 + RTX 4060 Ti 16GB mixed pair ⚠ uncheckedRTX 3090 + RTX 4060 Ti 16GB40Llama 3.3 70B: 11
RTX 3090: 2026 models (llamabench community)RTX 309024Qwen3.6 35B-A3B: 146-153
Qwen3.6 27B: 40.2
Qwen3.8 27B: 38.2-71.6
RTX 4060 Ti 16GB: 2026 models (llamabench community)RTX 4060 Ti 16GB16LFM2.5 24B-A2B: 131-150
Qwen3.8 27B: 10.7-17.6
RTX 4090 (llm-speed signed run, Oct 2026): Flint 40B-A2.5B at 169-179 tok/s on llama.cppRTX 409024Flint 40B-A2.5B: 179.4
Flint 40B-A2.5B: 169.2
RTX 4090 (llm-speed signed runs, Oct 2026): Qwen3.6 35B-A3B 161 tok/s, Gemma 4 12B QAT 98 tok/sRTX 409024Qwen3.6 35B-A3B: 160.6
Gemma 4 12B: 97.7
GLM 4.7 Flash: 164.9
RTX 4090 desktop (speed comparison vs Macs)RTX 4090 24GB24Llama 3.1 8B: 52.99
Mistral Small 22B: 26.72
RTX 4090: 2026 models (llamabench community)RTX 409024Qwen3.6 27B: 47.0
Gemma 4 31B: 43.8
Gemma 4 26B-A4B: 164
Gemma 4 12B: 98.1
gpt-oss 20B: 203
Qwen3.5 9B: 137
LFM2.5 24B-A2B: 261-339
Nemotron 3.5 Bolt 74B-A9B: 108
RTX 5070 12GB + 32 GB DDR4: Qwen3.8-Flash-Next 177B at 11.5 tok/s with expert streamingRTX 5070 12GB12Qwen3.8 Flash-Next: 11.5
Qwen3.8 Flash-Next: 10.15
RTX 5070 Ti: 2026 models (llamabench community)RTX 5070 Ti16Gemma 4 12B: 87.6
RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/sRTX 509032gpt-oss 20B: 324.2
Qwen3.6 27B: 75.1
Qwen3.8 27B: 75.2
RTX 5090 workstation: Qwen3.8 Flash-Next IQ3_S at 159.2 tok/s on Strata (expert offload)RTX 509032Qwen3.8 Flash-Next: 159.2
RTX 5090: 2026 models (llamabench community)RTX 509032Qwen3.6 35B-A3B: 258
Qwen3.6 27B: 75.7
Gemma 4 31B: 70.0
Gemma 4 26B-A4B: 235
Gemma 4 12B: 142
gpt-oss 20B: 313
Qwen3.5 9B: 208
RTX A4000 16GB (llm-speed signed run, Oct 2026): Muse Glimmer 30B UD-IQ3_XXS at 21.4 tok/sRTX A400016Muse Glimmer 30B: 21.42
RTX Pro 6000 96GB workstation cardRTX Pro 6000 (Blackwell)96gpt-oss 120B: 196.31
gpt-oss 20B: 286.91
RTX Pro 6000 Blackwell Max-Q (300 W cap): Qwen3.8 27B on vLLM, NVFP4 vs BF16RTX Pro 6000 Max-Q (Blackwell, 96 GB)96Qwen3.8 27B: 150 ⚡
Qwen3.8 27B: 85 ⚡
Ryzen 9 7900X + RTX 4090RTX 409024Llama 3.2 3B: 237.05
Llama 3.1 70B: 3.1
Ryzen 9 7950X + RTX 4070 Ti SuperRTX 4070 Ti Super16Llama 3.2 3B: 198.95
Ryzen 9 7950X + RTX 4080RTX 408016Llama 3.2 3B: 204.45
Ryzen 9 9950X + Radeon RX 7900 XTAMD Radeon RX 7900 XT20Llama 3.2 3B: 131.2
Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s)RTX 509032Qwen3.8 Flash-Next: 78.8 ⚡
Ornith 1.5 122B: 46.4
GLM-5.3 Flash: 30.1 ⚡
DeepSeek V4.1 Flash: 24.9
Single Intel Arc Pro B70 32 GB (neural.download lab)Intel Arc Pro B70 32 GB32Gemma 4 26B-A4B: 122.16
Ornith 1.5 35B-A3B: 131.46
Qwen3.8 27B: 27.8
Single Intel Arc Pro B70 with vLLM XPU cookbookIntel Arc Pro B70 32 GB32Qwen3.8 27B: 96 ⚡
Qwen3.8 27B: 66 ⚡
Qwen3.6 35B-A3B: 120
Single RTX 3090 24GBRTX 309024gpt-oss 20B: 161.77
Single RTX 3090: Qwen3.6 35B-A3B (Giles Thomas)RTX 309024Qwen3.6 35B-A3B: 139.6
Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cppRTX 309024Qwen3.8 27B: 139.7 ⚡
Qwen3.8 27B: 72.5 ⚡
Qwen3.8 27B: 78.0 ⚡
Qwen3.8 27B: 50.9 ⚡
Qwen3.8 27B: 40.4
Qwen3.8 27B: 42.9
Single RTX 4060 Ti 16GB + 64 GB RAM (partial offload) ⚠ uncheckedRTX 4060 Ti 16GB16Llama 3.3 70B: 1.5
Single RTX 4090 (llama.cpp CUDA scoreboard)RTX 4090 24GB24Llama 2 7B: 188.96
Single RTX 5060 Ti 16GBRTX 5060 Ti 16GB16gpt-oss 20B: 111.51
Single RTX 5070 Ti 16GBRTX 5070 Ti16gpt-oss 20B: 189.45
Single RTX 5080 16GBRTX 508016gpt-oss 20B: 204.85
Single RTX 5090 32GBRTX 509032gpt-oss 20B: 282.51
Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3BAMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)128Qwen3.6 35B-A3B: 54.22
Qwen3.6 35B-A3B: 47.47
Qwen3.6 35B-A3B: 60.92 ⚡
Qwen3.8 27B: 7.76
Qwen3.8 27B: 7.77
Qwen3.8 27B: 15.67 ⚡
Qwen3.8 27B: 13.19 ⚡
Strix Halo (Radeon 8060S, 96 GB): 2026 models (llamabench community)AMD Ryzen AI Max+ (Radeon 8060S iGPU)96Qwen3.6 35B-A3B: 50.2-62.4
Strix Halo (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B Q6_K, 9.6 to 18 tok/s with MTPAMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)128Qwen3.8 27B: 9.59
Qwen3.8 27B: 18.01 ⚡
Qwen3.8 27B: 15.16 ⚡
Strix Halo + Radeon AI Pro R9700 over OCuLink: Qwen3.8 Flash-Next at ~102 tok/s (custom engine)AMD Radeon AI Pro R9700 32 GB (OCuLink x4) + Ryzen AI Max+ 395 Radeon 8060S iGPU160Qwen3.8 Flash-Next: 102 ⚡
Qwen3.8 Flash-Next: 82
Qwen3.8 Flash-Next: 71
Qwen3.8 Flash-Next: 29-34
Strix Halo mini PC (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B at 12.6 tok/s Q4_K_M, 7.7 Q8_0AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)128Qwen3.8 27B: 12.6
Qwen3.8 27B: 7.7