Radeon PRO W7700 (16GB) 提供 16 GB 显存/内存与 576 GB/s 带宽,支持通过 ROCm 与 Vulkan 加速进行本地推理。
总显存
16 GB
可用显存上限
16 GB
显存带宽
576 GB/s
最大推理模型级别
14B 模型 (Q4)
热设计功耗
190W
随上下文长度增长的显存需求评估。当曲线超过参考线时将导致显存不足(OOM)。
量化精度:
主流开源大模型在 Radeon PRO W7700 (16GB) 上的推理速度、显存兼容性与生成吞吐表现。
排序:
上下文长度:
KV 缓存:
显卡数量:
| 排名 | 模型 | 参数量 | 量化精度 | 预估 TPS | TTFT | 可运行 |
|---|---|---|---|---|---|---|
#51 | Inkling 2 Small | 2B (17B active) | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~3.2s ~3.2s ~3.3s | |
#55 | Qwen 3.8 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
#99 | Qwen3.5-27B | 27B | INT4 | 26 tok/s | ~6.2s | |
#115 | Phi-4 Reasoning Plus | 14B | INT4 | 37 tok/s | ~2.8s | |
#116 | MiMo V2 Flash | 15B (309B active) | INT4 | 3 tok/s | ~62.1s | |
#122 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 | 104 tok/s | ~1.0s | |
#129 | Qwen3.5-9B | 9B | INT4 Q8 | 70 tok/s 44 tok/s | ~1.8s ~1.8s | |
#141 | Gemma 4 12B | 11.95B | INT4 Q8 | 42 tok/s 26 tok/s | ~2.4s ~2.8s | |
#145 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
#150 | GPT-OSS 20B | 21B (3.6B active) | INT4 | 51 tok/s | ~1.0s | |
#153 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
#157 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 25 tok/s 15 tok/s 8 tok/s | ~5.6s ~5.6s ~5.7s | |
#164 | Granite 4.2 8B | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#167 | Gemma 4 E4B | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#172 | Phi-4 | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
#173 | Gemma 3 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
#185 | DeepSeek-R1 14B | 14B | INT4 Q8 | 51 tok/s 28 tok/s | ~2.8s ~3.3s | |
#189 | Ministral 3 14B | 14B | INT4 | 37 tok/s | ~2.8s | |
#192 | Gemma 3 12B | 12B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
#194 | Mistral-Small-2501 | 24B | INT4 | 30 tok/s | ~5.1s | |
#196 | Ministral 3 8B | 8B | INT4 Q8 | 57 tok/s 41 tok/s | ~1.6s ~1.6s | |
#197 | OLMo 3 7B Think | 7B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
#198 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
#199 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
#201 | Qwen2.5-14B | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
#204 | Gemma 2 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
#205 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 103 tok/s 70 tok/s 40 tok/s | ~1.0s ~1.0s ~1.1s | |
#209 | DeepSeek-R1 8B | 8B | INT4 Q8 | 79 tok/s 50 tok/s | ~1.6s ~1.6s | |
#211 | Qwen3-4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
#212 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 99 tok/s 79 tok/s 53 tok/s | ~634 ms ~636 ms ~642 ms | |
#214 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 220 tok/s 189 tok/s 145 tok/s | ~183 ms ~184 ms ~186 ms | |
#215 | LFM3 1.7B | 1.7B | INT4 Q8 FP16 | 175 tok/s 137 tok/s 93 tok/s | ~377 ms ~379 ms ~382 ms | |
#216 | Qwen3-14B | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
#217 | Qwen3-8B | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#218 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
#219 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 175 tok/s 137 tok/s 93 tok/s | ~377 ms ~379 ms ~382 ms | |
#220 | Gemma 2 9B | 9B | INT4 Q8 | 70 tok/s 44 tok/s | ~1.8s ~1.8s | |
#221 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 244 tok/s 231 tok/s 205 tok/s | ~66 ms ~66 ms ~67 ms | |
#222 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 86 tok/s 55 tok/s 29 tok/s | ~1.4s ~1.4s ~1.5s | |
#223 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 234 tok/s 207 tok/s 165 tok/s | ~137 ms ~138 ms ~139 ms | |
#225 | Llama 3.1 8B | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#228 | Llama 3 8B | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#229 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 122 tok/s 86 tok/s 51 tok/s | ~785 ms ~788 ms ~796 ms | |
#231 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
#232 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 183 tok/s 146 tok/s 101 tok/s | ~335 ms ~336 ms ~339 ms | |
#233 | LFM3 VL 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
#234 | Ministral-8B-2410 | 8B | INT4 Q8 | 77 tok/s 49 tok/s | ~1.6s ~1.6s | |
#235 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
#237 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 122 tok/s 86 tok/s 51 tok/s | ~785 ms ~788 ms ~796 ms | |
#238 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 99 tok/s 86 tok/s 60 tok/s | ~473 ms ~475 ms ~519 ms | |
#239 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 82 tok/s 53 tok/s 27 tok/s | ~1.5s ~1.5s ~1.7s | |
#240 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 207 tok/s 174 tok/s 128 tok/s | ~231 ms ~232 ms ~234 ms | |
#241 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
#242 | Phi-3-medium | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
#243 | OLMo 3 7B Instruct | 7B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
#245 | Qwen2-7B | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
#246 | Phi-3-small | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
#247 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 207 tok/s 174 tok/s 128 tok/s | ~231 ms ~232 ms ~234 ms | |
#248 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 82 tok/s 53 tok/s 27 tok/s | ~1.5s ~1.5s ~1.7s | |
#250 | Gemma 1 7B | 7B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
#251 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 172 tok/s 130 tok/s 85 tok/s | ~440 ms ~442 ms ~446 ms | |
#252 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 69 tok/s 51 tok/s 30 tok/s | ~1.2s ~1.2s ~1.3s | |
#253 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 69 tok/s 51 tok/s 30 tok/s | ~1.2s ~1.2s ~1.3s | |
#254 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 69 tok/s 51 tok/s 30 tok/s | ~1.2s ~1.2s ~1.3s | |
未上榜模型(按推理速度排序) | ||||||
- | SmolVLM2 256M Video Instruct | 0.26B | INT4 Q8 FP16 | 261 tok/s 246 tok/s 218 tok/s | ~64 ms ~64 ms ~64 ms | |
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 258 tok/s 241 tok/s 210 tok/s | ~72 ms ~72 ms ~73 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 242 tok/s 218 tok/s 178 tok/s | ~115 ms ~116 ms ~117 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 242 tok/s 218 tok/s 178 tok/s | ~115 ms ~116 ms ~117 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 239 tok/s 225 tok/s 198 tok/s | ~72 ms ~73 ms ~73 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 239 tok/s 225 tok/s 198 tok/s | ~72 ms ~73 ms ~73 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 234 tok/s 207 tok/s 165 tok/s | ~137 ms ~138 ms ~139 ms | |
- | PixArt-α XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 234 tok/s 206 tok/s 164 tok/s | ~140 ms ~140 ms ~141 ms | |
- | PixArt-Σ XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 234 tok/s 206 tok/s 164 tok/s | ~140 ms ~140 ms ~141 ms | |
- | Nemotron 3.5 Parse | 0.63B | INT4 Q8 FP16 | 232 tok/s 204 tok/s 162 tok/s | ~145 ms ~145 ms ~147 ms | |
- | SD 2.1 Turbo | 0.86B | INT4 Q8 FP16 | 216 tok/s 185 tok/s 139 tok/s | ~197 ms ~198 ms ~200 ms | |
- | Stable Diffusion 1.4 | 0.86B | INT4 Q8 FP16 | 216 tok/s 185 tok/s 139 tok/s | ~197 ms ~198 ms ~200 ms | |
- | Stable Diffusion 2.1 | 0.86B | INT4 Q8 FP16 | 216 tok/s 185 tok/s 139 tok/s | ~197 ms ~198 ms ~200 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 214 tok/s 179 tok/s 131 tok/s | ~231 ms ~232 ms ~234 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 207 tok/s 174 tok/s 128 tok/s | ~231 ms ~232 ms ~234 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 207 tok/s 174 tok/s 128 tok/s | ~231 ms ~232 ms ~234 ms | |
- | Kandinsky 2.2 Decoder | 1.22B | INT4 Q8 FP16 | 196 tok/s 161 tok/s 115 tok/s | ~277 ms ~278 ms ~280 ms | |
- | Wan 2.1 T2V 1.3B | 1.3B | INT4 Q8 FP16 | 192 tok/s 156 tok/s 110 tok/s | ~292 ms ~294 ms ~296 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 183 tok/s 146 tok/s 101 tok/s | ~335 ms ~336 ms ~339 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 183 tok/s 146 tok/s 101 tok/s | ~335 ms ~336 ms ~339 ms | |
- | SVD Image-to-Video XT | 1.5B | INT4 Q8 FP16 | 183 tok/s 146 tok/s 101 tok/s | ~335 ms ~336 ms ~339 ms | |
- | SVD Image-to-Video XT 1.1 | 1.5B | INT4 Q8 FP16 | 183 tok/s 146 tok/s 101 tok/s | ~335 ms ~336 ms ~339 ms | |
- | LFM3 1.7B Base | 1.7B | INT4 Q8 FP16 | 175 tok/s 137 tok/s 93 tok/s | ~377 ms ~379 ms ~382 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 175 tok/s 137 tok/s 93 tok/s | ~377 ms ~379 ms ~382 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~397 ms ~399 ms ~402 ms | |
- | CogVideoX 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | LFM3 Audio 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | LFM3 Omni 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | Liquid D1 2B | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | LTX-Video 0.9.1 | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | Moondream 3.1 | 9B (2B active) | INT4 Q8 | 164 tok/s 126 tok/s | ~495 ms ~497 ms | |
- | Stable Diffusion 3 Medium | 2B | INT4 Q8 FP16 | 164 tok/s 126 tok/s 83 tok/s | ~440 ms ~442 ms ~446 ms | |
- | Kandinsky 3.0 | 2.1B | INT4 Q8 FP16 | 161 tok/s 123 tok/s 80 tok/s | ~459 ms ~461 ms ~465 ms | |
- | LTX-2 | 2.2B | INT4 Q8 FP16 | 158 tok/s 120 tok/s 78 tok/s | ~478 ms ~480 ms ~485 ms | |
- | SDXL 1.0 Refiner | 2.31B | INT4 Q8 FP16 | 155 tok/s 117 tok/s 75 tok/s | ~499 ms ~501 ms ~506 ms | |
- | Stable Diffusion 3.5 Medium | 2.5B | INT4 Q8 FP16 | 150 tok/s 112 tok/s 71 tok/s | ~536 ms ~538 ms ~543 ms | |
- | SDXL 1.0 Base | 2.56B | INT4 Q8 FP16 | 148 tok/s 110 tok/s 70 tok/s | ~547 ms ~549 ms ~554 ms | |
- | SDXL Turbo | 2.56B | INT4 Q8 FP16 | 148 tok/s 110 tok/s 70 tok/s | ~547 ms ~549 ms ~554 ms | |
- | Alpamayo2 Edge | 2.65B | INT4 Q8 FP16 | 146 tok/s 108 tok/s 68 tok/s | ~564 ms ~567 ms ~572 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 144 tok/s 103 tok/s 63 tok/s | ~630 ms ~633 ms ~639 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 142 tok/s 124 tok/s 93 tok/s | ~294 ms ~295 ms ~298 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 142 tok/s 124 tok/s 93 tok/s | ~294 ms ~295 ms ~298 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 142 tok/s 124 tok/s 93 tok/s | ~294 ms ~295 ms ~298 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 140 tok/s 101 tok/s 62 tok/s | ~631 ms ~633 ms ~640 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 | 140 tok/s 101 tok/s | ~616 ms ~671 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 138 tok/s 100 tok/s 62 tok/s | ~631 ms ~634 ms ~640 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 | 124 tok/s | ~758 ms | |
- | FLUX.2 Klein 4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
- | Mage-VL Cowork | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 119 tok/s 83 tok/s 49 tok/s | ~823 ms ~827 ms ~835 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 | 118 tok/s | ~836 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 | 118 tok/s | ~836 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 115 tok/s 97 tok/s 69 tok/s | ~443 ms ~444 ms ~448 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 | 108 tok/s | ~950 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 | 108 tok/s | ~950 ms | |
- | CogVideoX 1.5 5B | 5B | INT4 Q8 FP16 | 105 tok/s 71 tok/s 41 tok/s | ~1.0s ~1.0s ~1.0s | |
- | CogVideoX 5B | 5B | INT4 Q8 FP16 | 105 tok/s 71 tok/s 41 tok/s | ~1.0s ~1.0s ~1.0s | |
- | Wan 2.2 TI2V 5B | 5B | INT4 Q8 FP16 | 105 tok/s 71 tok/s 41 tok/s | ~1.0s ~1.0s ~1.0s | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 103 tok/s 84 tok/s 57 tok/s | ~576 ms ~579 ms ~584 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 99 tok/s 79 tok/s 53 tok/s | ~634 ms ~636 ms ~642 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 99 tok/s 79 tok/s 53 tok/s | ~634 ms ~636 ms ~642 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 88 tok/s 56 tok/s 29 tok/s | ~1.4s ~1.4s ~1.5s | |
- | AuraFlow v0.2 | 6.8B | INT4 Q8 FP16 | 86 tok/s 56 tok/s 30 tok/s | ~1.4s ~1.4s ~1.5s | |
- | AuraFlow v0.3 | 6.8B | INT4 Q8 FP16 | 86 tok/s 56 tok/s 30 tok/s | ~1.4s ~1.4s ~1.5s | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
- | FastFrames-L | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 84 tok/s 55 tok/s 28 tok/s | ~1.4s ~1.4s ~1.7s | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 82 tok/s 53 tok/s 27 tok/s | ~1.5s ~1.5s ~1.7s | |
- | Youtu Parsing Omni | 8B | INT4 Q8 | 79 tok/s 50 tok/s | ~1.6s ~1.6s | |
- | Stable Diffusion 3.5 Large | 8.1B | INT4 Q8 | 76 tok/s 48 tok/s | ~1.6s ~1.6s | |
- | Stable Diffusion 3.5 Large Turbo | 8.1B | INT4 Q8 | 76 tok/s 48 tok/s | ~1.6s ~1.6s | |
- | FLUX.2 Klein 9B | 9B | INT4 Q8 | 70 tok/s 44 tok/s | ~1.8s ~1.8s | |
- | LensVLM 9B | 9B | INT4 Q8 | 70 tok/s 44 tok/s | ~1.8s ~1.8s | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 | 70 tok/s 44 tok/s | ~1.8s ~1.8s | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 69 tok/s 51 tok/s 30 tok/s | ~1.2s ~1.2s ~1.3s | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 69 tok/s 51 tok/s 30 tok/s | ~1.2s ~1.2s ~1.3s | |
- | Falcon3-10B | 10B | INT4 Q8 | 65 tok/s 41 tok/s | ~2.0s ~2.0s | |
- | Mochi 1 Preview | 10B | INT4 Q8 | 65 tok/s 41 tok/s | ~2.0s ~2.0s | |
- | Falcon2-11B | 11B | INT4 Q8 | 63 tok/s 38 tok/s | ~2.2s ~2.2s | |
- | Hunyuan Lite | 7B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 | 62 tok/s | ~810 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
- | SEA-LION-7B | 7.1B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 | 62 tok/s 45 tok/s | ~1.4s ~1.4s | |
- | FLUX.1 Canny Dev | 11.9B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
- | FLUX.1 Depth Dev | 11.9B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
- | FLUX.1 Dev | 11.9B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
- | FLUX.1 Fill Dev | 11.9B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
- | FLUX.1 Schnell | 11.9B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.4s ~2.6s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 | 57 tok/s 41 tok/s | ~1.6s ~1.6s | |
- | Typhoon-2-8B | 8B | INT4 Q8 | 57 tok/s 41 tok/s | ~1.6s ~1.6s | |
- | HunyuanVideo 1.5 720P T2V | 13B | INT4 Q8 | 53 tok/s 31 tok/s | ~2.6s ~2.8s | |
- | GLM-4-9B | 9B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | GLM-4V | 9B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | Yi-9B | 9B | INT4 Q8 | 52 tok/s 35 tok/s | ~1.8s ~2.0s | |
- | Wan 2.1 I2V 14B 720P | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
- | Wan 2.1 T2V 14B | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
- | Wan 2.2 T2V A14B | 14B | INT4 Q8 | 50 tok/s 27 tok/s | ~2.8s ~3.3s | |
- | FLUX.2 Dev | 15B | INT4 Q8 | 47 tok/s 26 tok/s | ~2.9s ~3.5s | |
- | GreenMind-14B-R1 | 14B | INT4 | 37 tok/s | ~2.8s | |
- | Arete 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
- | AREX-2 | 27B | INT4 | 26 tok/s | ~6.2s | |
- | PPLX Decider v1 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
- | PPLX Decider v1.1 27B | 27B | INT4 | 26 tok/s | ~6.2s | |
- | PPLX Decider v1.3 | 27B | INT4 | 26 tok/s | ~6.2s | |
不同量化精度下可运行的最大模型参数级别。
上下文长度:
4位量化
14B 模型
8位精度
8B 模型
16位全精度
3B 模型
本地模型训练与适配器微调支持能力。
微调上下文:
QLoRA
14B 模型
LoRA
8B 模型
全参数微调
1B 模型
Radeon PRO W7700 (16GB) 上的最佳量化精度与运行规格建议。
最佳推理区间
非常适合 8B 模型长上下文推理(32k+)。14B 模型在 Q4 精度及标准上下文下也能流畅运行。
显存带宽与推理速率
具备 576 GB/s 聚合带宽,单批次推理处于显存带宽受限模式。8B Q4 模型预估生成速度约 128 tok/s,70B Q4 模型预估生成速度约 15 tok/s。