AMD Ryzen AI MAX+ 395 (128GB system, 96GB GPU-accessible)
The AMD Ryzen AI MAX+ 395 (128GB system, 96GB GPU-accessible) provides 96 GB of memory with 256 GB/s bandwidth, supporting local inference via ROCm and Vulkan acceleration.
Memory
96 GB
Usable Memory Ceiling
96 GB
Bandwidth
256 GB/s
Max Inference Class
70B+ Models (Q4)
TDP
120W
VRAM Scaling by Context Length
Estimated memory requirements as context length increases. Curves crossing above the reference line indicate Out of Memory (OOM).
Precision:
Best Models to Run on AMD Ryzen AI MAX+ 395 (128GB system, 96GB GPU-accessible)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on AMD Ryzen AI MAX+ 395 (128GB system, 96GB GPU-accessible).
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#42 | Qwen 3.8 Flash | 176B (6B active) | INT4 | 50 tok/s | ~1.3s | |
#48 | Qwen 3.8 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
#72 | Sarvam-105B | 106B (10.3B active) | INT4 | 32 tok/s | ~2.9s | |
#83 | K2 Horizon MoVA 36B A4B | 36B (4B active) | INT4 Q8 FP16 | 58 tok/s 39 tok/s 23 tok/s | ~1.1s ~1.1s ~1.1s | |
#88 | Ling 3.0 Flash VL | 124B (5.5B active) | INT4 | 51 tok/s | ~2.1s | |
#93 | Qwen3.5-27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
#97 | Qwen3.5-122B-A10B | 122B (10B active) | INT4 | 25 tok/s | ~3.0s | |
#105 | Sarvam-30B | 32B (2.4B active) | INT4 Q8 FP16 | 76 tok/s 56 tok/s 35 tok/s | ~779 ms ~784 ms ~795 ms | |
#106 | Qwen3.6 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 68 tok/s 48 tok/s 29 tok/s | ~908 ms ~914 ms ~928 ms | |
#107 | Nemotron 3.5 Bolt 74B A9B | 74B | INT4 Q8 | 5 tok/s 3 tok/s | ~15.4s ~16.9s | |
#110 | Muse Glimmer 30B | 30B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 4 tok/s | ~6.3s ~6.4s ~6.5s | |
#111 | MiMo V2 Flash | 15B (309B active) | INT4 Q8 FP16 | 1 tok/s 1 tok/s 1 tok/s | ~62.0s ~62.7s ~64.1s | |
#112 | Phi-4 Reasoning Plus | 14B | INT4 Q8 FP16 | 18 tok/s 12 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#117 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 Q8 FP16 | 62 tok/s 42 tok/s 24 tok/s | ~1.0s ~1.0s ~1.0s | |
#119 | GLM-4.5-Air | 106B (12B active) | INT4 | 9 tok/s | ~3.3s | |
#124 | Ling 3.0 Flash | 124B (5.1B active) | INT4 | 54 tok/s | ~2.0s | |
#125 | Qwen3.5-9B | 9B | INT4 Q8 FP16 | 36 tok/s 22 tok/s 11 tok/s | ~1.9s ~2.0s ~2.0s | |
#126 | Qwen3.5-35B-A3B | 35B (3B active) | INT4 Q8 FP16 | 68 tok/s 48 tok/s 29 tok/s | ~908 ms ~914 ms ~928 ms | |
#128 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
#135 | Gemma 4 12B | 11.95B | INT4 Q8 FP16 | 20 tok/s 14 tok/s 8 tok/s | ~2.6s ~2.6s ~2.6s | |
#137 | Qwen3 Next 80B A3B | 80B (79B active) | INT4 Q8 | 4 tok/s 2 tok/s | ~16.5s ~19.5s | |
#140 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
#143 | GPT-OSS 20B | 21B (3.6B active) | INT4 Q8 FP16 | 29 tok/s 25 tok/s 17 tok/s | ~974 ms ~980 ms ~996 ms | |
#144 | GLM-4V | 9B | INT4 Q8 FP16 | 26 tok/s 18 tok/s 10 tok/s | ~1.9s ~2.0s ~2.0s | |
#147 | GPT-OSS 120B | 117B (5.1B active) | INT4 | 9 tok/s | ~2.2s | |
#150 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 12 tok/s 7 tok/s 3 tok/s | ~6.1s ~6.2s ~6.3s | |
#153 | Qwen3-30B-A3B | 30B (3B active) | INT4 Q8 FP16 | 69 tok/s 49 tok/s 29 tok/s | ~891 ms ~897 ms ~911 ms | |
#155 | Mistral Medium 3.5 128B | 128B | INT4 | 2 tok/s | ~29.0s | |
#158 | Granite 4.2 8B | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#159 | Gemma 4 E4B | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#162 | Qwen2.5-72B | 72B | INT4 Q8 | 5 tok/s 3 tok/s | ~15.0s ~15.1s | |
#163 | Mistral Large 3 | 41B (675B active) | INT4 Q8 FP16 | 1 tok/s 1 tok/s 1 tok/s | ~135.4s ~136.8s ~164.1s | |
#164 | DeepSeek-R1 70B | 70B | INT4 Q8 | 6 tok/s 3 tok/s | ~14.5s ~14.7s | |
#165 | GLM-4 | 32B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#166 | Phi-4 | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#167 | Gemma 3 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
#168 | Llama 3.3 70B | 70B | INT4 Q8 | 5 tok/s 3 tok/s | ~14.5s ~14.7s | |
#170 | DeepSeek-R1 32B | 32B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#171 | Qwen3-32B | 32B | INT4 Q8 FP16 | 11 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#172 | Nemotron 3.5 Super VL | 120B | INT4 | 3 tok/s | ~25.0s | |
#173 | Hunyuan Standard | 52B (389B active) | INT4 Q8 | 1 tok/s 1 tok/s | ~78.2s ~79.0s | |
#175 | Magistral Small | 24B | INT4 Q8 FP16 | 11 tok/s 7 tok/s 4 tok/s | ~5.1s ~5.1s ~5.2s | |
#176 | OLMo 3.1 32B Instruct | 32B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#177 | Qwen2.5-32B | 32B | INT4 Q8 FP16 | 11 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#178 | OLMo 3 7B Think | 7B | INT4 Q8 FP16 | 32 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.5s ~1.6s | |
#179 | DeepSeek-R1 14B | 14B | INT4 Q8 FP16 | 25 tok/s 15 tok/s 8 tok/s | ~3.0s ~3.0s ~3.1s | |
#180 | Mistral-Large-2407 | 123B | INT4 | 3 tok/s | ~25.6s | |
#181 | Llama 3.1 70B | 70B | INT4 Q8 | 5 tok/s 3 tok/s | ~14.5s ~14.7s | |
#183 | Ministral 3 14B | 14B | INT4 Q8 FP16 | 18 tok/s 12 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#184 | Llama 4 Scout | 109B (17B active) | INT4 | 18 tok/s | ~4.3s | |
#185 | Command A | 111B | INT4 | 3 tok/s | ~23.2s | |
#186 | Gemma 3 12B | 12B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.6s ~2.6s ~2.6s | |
#187 | Qwen2-72B | 72B | INT4 Q8 | 5 tok/s 3 tok/s | ~15.0s ~15.1s | |
#188 | Mistral-Small-2501 | 24B | INT4 Q8 FP16 | 15 tok/s 9 tok/s 4 tok/s | ~5.1s ~5.1s ~5.2s | |
#189 | Ministral 3 8B | 8B | INT4 Q8 FP16 | 29 tok/s 20 tok/s 11 tok/s | ~1.7s ~1.8s ~1.8s | |
#190 | Gemma 4 31B | 30.7B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 3 tok/s | ~6.4s ~6.5s ~6.6s | |
#191 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
#192 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
#193 | Gemma 2 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
#194 | Qwen2.5-14B | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#195 | OLMo 3.1 32B Think | 32B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#196 | Nemotron 3.5 Lightning | 30B (3B active) | INT4 Q8 FP16 | 26 tok/s 23 tok/s 18 tok/s | ~916 ms ~920 ms ~933 ms | |
#198 | Llama 3 70B | 70B | INT4 Q8 | 5 tok/s 3 tok/s | ~14.5s ~14.7s | |
#199 | Command R Plus | 104B | INT4 | 3 tok/s | ~21.7s | |
#200 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 58 tok/s 36 tok/s 19 tok/s | ~1.1s ~1.1s ~1.2s | |
#202 | DeepSeek-R1 8B | 8B | INT4 Q8 FP16 | 41 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#203 | Kimi Linear 48B A3B Instruct | 48B (3B active) | INT4 Q8 | 22 tok/s 20 tok/s | ~993 ms ~996 ms | |
#204 | Qwen3-4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
#205 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 54 tok/s 42 tok/s 26 tok/s | ~690 ms ~695 ms ~709 ms | |
#206 | OLMo 3 32B Base | 32B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~6.7s ~6.8s ~6.9s | |
#207 | Gemma 2 9B | 9B | INT4 Q8 FP16 | 36 tok/s 22 tok/s 11 tok/s | ~1.9s ~2.0s ~2.0s | |
#208 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 171 tok/s 133 tok/s 89 tok/s | ~198 ms ~200 ms ~203 ms | |
#209 | Qwen3-14B | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#210 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
#211 | Qwen3-8B | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#212 | Mixtral-8x22B-v0.1 | 176B (22B active) | INT4 | 13 tok/s | ~6.3s | |
#213 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 46 tok/s 28 tok/s 15 tok/s | ~1.5s ~1.5s ~1.6s | |
#214 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 118 tok/s 83 tok/s 50 tok/s | ~409 ms ~412 ms ~420 ms | |
#215 | Llama 3.1 8B | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#216 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 206 tok/s 186 tok/s 151 tok/s | ~71 ms ~71 ms ~73 ms | |
#217 | Command R | 35B | INT4 Q8 FP16 | 8 tok/s 5 tok/s 3 tok/s | ~7.3s ~7.4s ~8.2s | |
#218 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 191 tok/s 154 tok/s 108 tok/s | ~148 ms ~150 ms ~152 ms | |
#219 | Llama 3 8B | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#221 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 71 tok/s 46 tok/s 25 tok/s | ~850 ms ~858 ms ~875 ms | |
#222 | Mixtral-8x7B-v0.1 | 46.7B (7B active) | INT4 Q8 FP16 | 39 tok/s 25 tok/s 13 tok/s | ~1.8s ~1.8s ~2.1s | |
#223 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
#224 | Ministral-8B-2410 | 8B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
#225 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 126 tok/s 91 tok/s 56 tok/s | ~362 ms ~365 ms ~372 ms | |
#226 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
#227 | Nemotron 3.5 Lightning 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 69 tok/s 49 tok/s 29 tok/s | ~891 ms ~897 ms ~911 ms | |
#228 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 71 tok/s 46 tok/s 25 tok/s | ~850 ms ~858 ms ~875 ms | |
#229 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 54 tok/s 46 tok/s 32 tok/s | ~517 ms ~520 ms ~530 ms | |
#230 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 43 tok/s 26 tok/s 14 tok/s | ~1.6s ~1.6s ~1.6s | |
#231 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 154 tok/s 117 tok/s 76 tok/s | ~250 ms ~252 ms ~256 ms | |
#232 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
#233 | Phi-3-medium | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
#234 | OLMo 3 7B Instruct | 7B | INT4 Q8 FP16 | 32 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.5s ~1.6s | |
#235 | Yi-34B | 34B | INT4 Q8 FP16 | 8 tok/s 5 tok/s 3 tok/s | ~7.1s ~7.2s ~7.4s | |
#236 | Qwen2-7B | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
#237 | Phi-3-small | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
#238 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 154 tok/s 117 tok/s 76 tok/s | ~250 ms ~252 ms ~256 ms | |
#239 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 43 tok/s 26 tok/s 14 tok/s | ~1.6s ~1.6s ~1.6s | |
#240 | Falcon-180B | 180B | INT4 | 2 tok/s | ~43.5s | |
#241 | Gemma 1 7B | 7B | INT4 Q8 FP16 | 32 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.5s ~1.6s | |
#242 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 115 tok/s 78 tok/s 45 tok/s | ~476 ms ~480 ms ~489 ms | |
#243 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 35 tok/s 25 tok/s 15 tok/s | ~1.3s ~1.3s ~1.4s | |
#244 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 35 tok/s 25 tok/s 15 tok/s | ~1.3s ~1.3s ~1.4s | |
#245 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 35 tok/s 25 tok/s 15 tok/s | ~1.3s ~1.3s ~1.4s | |
Unranked Models (Sorted by Speed) | ||||||
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 230 tok/s 201 tok/s 158 tok/s | ~77 ms ~78 ms ~79 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 203 tok/s 168 tok/s 121 tok/s | ~124 ms ~125 ms ~128 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 203 tok/s 168 tok/s 121 tok/s | ~124 ms ~125 ms ~128 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 198 tok/s 178 tok/s 143 tok/s | ~78 ms ~79 ms ~80 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 198 tok/s 178 tok/s 143 tok/s | ~78 ms ~79 ms ~80 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 191 tok/s 154 tok/s 108 tok/s | ~148 ms ~150 ms ~152 ms | |
- | PixArt-α XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 190 tok/s 153 tok/s 107 tok/s | ~151 ms ~152 ms ~155 ms | |
- | PixArt-Σ XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 190 tok/s 153 tok/s 107 tok/s | ~151 ms ~152 ms ~155 ms | |
- | Nemotron 3.5 Parse | 0.63B | INT4 Q8 FP16 | 188 tok/s 151 tok/s 105 tok/s | ~156 ms ~157 ms ~160 ms | |
- | SD 2.1 Turbo | 0.86B | INT4 Q8 FP16 | 166 tok/s 128 tok/s 85 tok/s | ~213 ms ~215 ms ~219 ms | |
- | Stable Diffusion 1.4 | 0.86B | INT4 Q8 FP16 | 166 tok/s 128 tok/s 85 tok/s | ~213 ms ~215 ms ~219 ms | |
- | Stable Diffusion 2.1 | 0.86B | INT4 Q8 FP16 | 166 tok/s 128 tok/s 85 tok/s | ~213 ms ~215 ms ~219 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 163 tok/s 122 tok/s 78 tok/s | ~249 ms ~251 ms ~256 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 154 tok/s 117 tok/s 76 tok/s | ~250 ms ~252 ms ~256 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 154 tok/s 117 tok/s 76 tok/s | ~250 ms ~252 ms ~256 ms | |
- | Kandinsky 2.2 Decoder | 1.22B | INT4 Q8 FP16 | 140 tok/s 104 tok/s 66 tok/s | ~299 ms ~302 ms ~308 ms | |
- | Wan 2.1 T2V 1.3B | 1.3B | INT4 Q8 FP16 | 136 tok/s 100 tok/s 63 tok/s | ~316 ms ~319 ms ~325 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 126 tok/s 91 tok/s 56 tok/s | ~362 ms ~365 ms ~372 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 126 tok/s 91 tok/s 56 tok/s | ~362 ms ~365 ms ~372 ms | |
- | SVD Image-to-Video XT | 1.5B | INT4 Q8 FP16 | 126 tok/s 91 tok/s 56 tok/s | ~362 ms ~365 ms ~372 ms | |
- | SVD Image-to-Video XT 1.1 | 1.5B | INT4 Q8 FP16 | 126 tok/s 91 tok/s 56 tok/s | ~362 ms ~365 ms ~372 ms | |
- | LFM3 1.7B | 1.7B | INT4 Q8 FP16 | 118 tok/s 83 tok/s 50 tok/s | ~409 ms ~412 ms ~420 ms | |
- | LFM3 1.7B Base | 1.7B | INT4 Q8 FP16 | 118 tok/s 83 tok/s 50 tok/s | ~409 ms ~412 ms ~420 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 118 tok/s 83 tok/s 50 tok/s | ~409 ms ~412 ms ~420 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 114 tok/s 80 tok/s 48 tok/s | ~430 ms ~434 ms ~442 ms | |
- | CogVideoX 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | LFM3 Audio 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | LFM3 Omni 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | LFM3 VL 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | Liquid D1 2B | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | LTX-Video 0.9.1 | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | Moondream 3.1 | 9B (2B active) | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~536 ms ~540 ms ~549 ms | |
- | Stable Diffusion 3 Medium | 2B | INT4 Q8 FP16 | 107 tok/s 74 tok/s 44 tok/s | ~477 ms ~481 ms ~490 ms | |
- | Kandinsky 3.0 | 2.1B | INT4 Q8 FP16 | 104 tok/s 72 tok/s 42 tok/s | ~497 ms ~502 ms ~511 ms | |
- | LTX-2 | 2.2B | INT4 Q8 FP16 | 102 tok/s 69 tok/s 41 tok/s | ~518 ms ~522 ms ~533 ms | |
- | SDXL 1.0 Refiner | 2.31B | INT4 Q8 FP16 | 99 tok/s 67 tok/s 39 tok/s | ~541 ms ~545 ms ~556 ms | |
- | Stable Diffusion 3.5 Medium | 2.5B | INT4 Q8 FP16 | 94 tok/s 63 tok/s 37 tok/s | ~580 ms ~585 ms ~597 ms | |
- | SDXL 1.0 Base | 2.56B | INT4 Q8 FP16 | 93 tok/s 62 tok/s 36 tok/s | ~592 ms ~598 ms ~609 ms | |
- | SDXL Turbo | 2.56B | INT4 Q8 FP16 | 93 tok/s 62 tok/s 36 tok/s | ~592 ms ~598 ms ~609 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 89 tok/s 57 tok/s 32 tok/s | ~683 ms ~689 ms ~703 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 87 tok/s 73 tok/s 51 tok/s | ~320 ms ~323 ms ~329 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 87 tok/s 73 tok/s 51 tok/s | ~320 ms ~323 ms ~329 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 87 tok/s 73 tok/s 51 tok/s | ~320 ms ~323 ms ~329 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 85 tok/s 56 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 FP16 | 85 tok/s 60 tok/s 35 tok/s | ~667 ms ~672 ms ~684 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 31 tok/s | ~683 ms ~689 ms ~703 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 Q8 FP16 | 78 tok/s 56 tok/s 34 tok/s | ~756 ms ~761 ms ~772 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 Q8 FP16 | 72 tok/s 50 tok/s 30 tok/s | ~834 ms ~840 ms ~854 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 Q8 FP16 | 72 tok/s 50 tok/s 30 tok/s | ~834 ms ~840 ms ~854 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 Q8 FP16 | 70 tok/s 49 tok/s 29 tok/s | ~877 ms ~883 ms ~897 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 Q8 FP16 | 70 tok/s 49 tok/s 29 tok/s | ~877 ms ~883 ms ~897 ms | |
- | FLUX.2 Klein 4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
- | Hy-Memory 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 69 tok/s 49 tok/s 29 tok/s | ~891 ms ~897 ms ~911 ms | |
- | Hy-MT2 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 69 tok/s 49 tok/s 29 tok/s | ~891 ms ~897 ms ~911 ms | |
- | Mage-VL Cowork | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 69 tok/s 44 tok/s 24 tok/s | ~892 ms ~900 ms ~918 ms | |
- | MiniCPM-V 4.7 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 68 tok/s 48 tok/s 29 tok/s | ~908 ms ~914 ms ~928 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 66 tok/s 53 tok/s 36 tok/s | ~482 ms ~486 ms ~495 ms | |
- | CogVideoX 1.5 5B | 5B | INT4 Q8 FP16 | 58 tok/s 36 tok/s 20 tok/s | ~1.1s ~1.1s ~1.1s | |
- | CogVideoX 5B | 5B | INT4 Q8 FP16 | 58 tok/s 36 tok/s 20 tok/s | ~1.1s ~1.1s ~1.1s | |
- | Wan 2.2 TI2V 5B | 5B | INT4 Q8 FP16 | 58 tok/s 36 tok/s 20 tok/s | ~1.1s ~1.1s ~1.1s | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 57 tok/s 44 tok/s 29 tok/s | ~628 ms ~633 ms ~645 ms | |
- | AliceAI Foundation 80B A3B Base | 80B (3B active) | INT4 Q8 | 54 tok/s 39 tok/s | ~1.2s ~1.3s | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 54 tok/s 42 tok/s 26 tok/s | ~690 ms ~695 ms ~709 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 54 tok/s 42 tok/s 26 tok/s | ~690 ms ~695 ms ~709 ms | |
- | Kolibri 1 | 78B (3.5B active) | INT4 Q8 | 51 tok/s 36 tok/s | ~1.3s ~1.4s | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 47 tok/s 28 tok/s 15 tok/s | ~1.5s ~1.5s ~1.6s | |
- | AuraFlow v0.2 | 6.8B | INT4 Q8 FP16 | 46 tok/s 28 tok/s 15 tok/s | ~1.5s ~1.5s ~1.5s | |
- | AuraFlow v0.3 | 6.8B | INT4 Q8 FP16 | 46 tok/s 28 tok/s 15 tok/s | ~1.5s ~1.5s ~1.5s | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 45 tok/s 27 tok/s 14 tok/s | ~1.5s ~1.5s ~1.6s | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 43 tok/s 26 tok/s 14 tok/s | ~1.6s ~1.6s ~1.6s | |
- | Youtu Parsing Omni | 8B | INT4 Q8 FP16 | 41 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Stable Diffusion 3.5 Large | 8.1B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.8s ~1.8s | |
- | Stable Diffusion 3.5 Large Turbo | 8.1B | INT4 Q8 FP16 | 40 tok/s 24 tok/s 13 tok/s | ~1.7s ~1.8s ~1.8s | |
- | FLUX.2 Klein 9B | 9B | INT4 Q8 FP16 | 36 tok/s 22 tok/s 11 tok/s | ~1.9s ~2.0s ~2.0s | |
- | LensVLM 9B | 9B | INT4 Q8 FP16 | 36 tok/s 22 tok/s 11 tok/s | ~1.9s ~2.0s ~2.0s | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 FP16 | 36 tok/s 22 tok/s 11 tok/s | ~1.9s ~2.0s ~2.0s | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 35 tok/s 25 tok/s 15 tok/s | ~1.3s ~1.3s ~1.4s | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 35 tok/s 25 tok/s 15 tok/s | ~1.3s ~1.3s ~1.4s | |
- | Falcon3-10B | 10B | INT4 Q8 FP16 | 33 tok/s 20 tok/s 10 tok/s | ~2.1s ~2.2s ~2.2s | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 Q8 FP16 | 33 tok/s 29 tok/s 21 tok/s | ~817 ms ~822 ms ~836 ms | |
- | Mochi 1 Preview | 10B | INT4 Q8 FP16 | 33 tok/s 20 tok/s 10 tok/s | ~2.1s ~2.2s ~2.2s | |
- | Falcon2-11B | 11B | INT4 Q8 FP16 | 32 tok/s 18 tok/s 10 tok/s | ~2.4s ~2.4s ~2.4s | |
- | Hunyuan Lite | 7B | INT4 Q8 FP16 | 32 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.5s ~1.6s | |
- | OLMo 3 7B Base | 7B | INT4 Q8 FP16 | 32 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.5s ~1.6s | |
- | SEA-LION-7B | 7.1B | INT4 Q8 FP16 | 31 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.6s ~1.6s | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 FP16 | 31 tok/s 22 tok/s 13 tok/s | ~1.5s ~1.6s ~1.6s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 FP16 | 29 tok/s 20 tok/s 11 tok/s | ~1.7s ~1.8s ~1.8s | |
- | Typhoon-2-8B | 8B | INT4 Q8 FP16 | 29 tok/s 20 tok/s 11 tok/s | ~1.7s ~1.8s ~1.8s | |
- | FLUX.1 Canny Dev | 11.9B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.5s ~2.6s ~2.6s | |
- | FLUX.1 Depth Dev | 11.9B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.5s ~2.6s ~2.6s | |
- | FLUX.1 Dev | 11.9B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.5s ~2.6s ~2.6s | |
- | FLUX.1 Fill Dev | 11.9B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.5s ~2.6s ~2.6s | |
- | FLUX.1 Schnell | 11.9B | INT4 Q8 FP16 | 28 tok/s 17 tok/s 9 tok/s | ~2.5s ~2.6s ~2.6s | |
- | Laguna S 2.1 | 118B (8B active) | INT4 | 28 tok/s | ~2.6s | |
- | GLM-4-9B | 9B | INT4 Q8 FP16 | 26 tok/s 18 tok/s 10 tok/s | ~1.9s ~2.0s ~2.0s | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 FP16 | 26 tok/s 18 tok/s 10 tok/s | ~1.9s ~2.0s ~2.0s | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 FP16 | 26 tok/s 18 tok/s 10 tok/s | ~1.9s ~2.0s ~2.0s | |
- | HunyuanVideo 1.5 720P T2V | 13B | INT4 Q8 FP16 | 26 tok/s 15 tok/s 8 tok/s | ~2.8s ~2.8s ~2.9s | |
- | Yi-9B | 9B | INT4 Q8 FP16 | 26 tok/s 18 tok/s 10 tok/s | ~1.9s ~2.0s ~2.0s | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 FP16 | 25 tok/s 17 tok/s 10 tok/s | ~2.0s ~2.0s ~2.0s | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 FP16 | 25 tok/s 17 tok/s 10 tok/s | ~2.0s ~2.0s ~2.0s | |
- | Wan 2.1 I2V 14B 720P | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
- | Wan 2.1 T2V 14B | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
- | Wan 2.2 T2V A14B | 14B | INT4 Q8 FP16 | 25 tok/s 14 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
- | FLUX.2 Dev | 15B | INT4 Q8 FP16 | 23 tok/s 13 tok/s 7 tok/s | ~3.2s ~3.2s ~3.3s | |
- | GreenMind-14B-R1 | 14B | INT4 Q8 FP16 | 18 tok/s 12 tok/s 7 tok/s | ~3.0s ~3.0s ~3.1s | |
- | Arete 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
- | AREX-2 | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
- | PPLX Decider v1 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
- | PPLX Decider v1.1 27B | 27B | INT4 Q8 FP16 | 13 tok/s 8 tok/s 4 tok/s | ~5.7s ~5.7s ~5.9s | |
- | Agnes 3.0 Flash | 33B | INT4 Q8 FP16 | 11 tok/s 6 tok/s 3 tok/s | ~6.9s ~7.0s ~7.1s | |
- | Falcon-40B | 40B | INT4 Q8 FP16 | 10 tok/s 5 tok/s 3 tok/s | ~8.3s ~8.4s ~9.3s | |
- | Hunyuan A13B | 80B (13B active) | INT4 Q8 | 10 tok/s 8 tok/s | ~3.3s ~3.9s | |
- | Kimi-Dev-72B | 72B | INT4 Q8 | 4 tok/s 2 tok/s | ~15.0s ~16.5s | |
- | Typhoon-2-70B | 70B | INT4 Q8 | 4 tok/s 3 tok/s | ~14.6s ~16.0s | |
- | GLM-130B | 130B | INT4 | 2 tok/s | ~29.4s | |
Inference Capacity
Maximum model parameter class runnable by weight precision.
Context:
4-bit
70B+ Models
8-bit
70B Models
16-bit
32B Models
Fine-Tuning Capacity
Hardware capacity for local training and adapter fine-tuning.
Context Length:
QLoRA
70B+ Models
LoRA
70B Models
Full Parameter Training
7B-8B Models
Workload Recommendations
Guidance for optimal precision and operational ceilings on AMD Ryzen AI MAX+ 395 (128GB system, 96GB GPU-accessible).
Inference Sweet Spot
Suitable for 70B parameter models at Q4/Q8 with extended 32k context windows, or high-throughput batching of smaller dense architectures.
Bandwidth & Speed Profile
With 256 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 57 tok/s on an 8B Q4 model and 7 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning