AMD RX 9070 (16GB)
The AMD RX 9070 (16GB) provides 16 GB of memory with 640 GB/s bandwidth, supporting local inference via ROCm and Vulkan acceleration.
Memory
16 GB
Usable Memory Ceiling
16 GB
Bandwidth
640 GB/s
Max Inference Class
14B Models (Q4)
TDP
220W
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 RX 9070 (16GB)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on AMD RX 9070 (16GB).
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#48 | Qwen 3.8 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
#93 | Qwen3.5-27B | 27B | INT4 | 28 tok/s | ~2.7s | |
#111 | MiMo V2 Flash | 15B (309B active) | INT4 | 3 tok/s | ~27.7s | |
#112 | Phi-4 Reasoning Plus | 14B | INT4 | 41 tok/s | ~1.2s | |
#117 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 | 111 tok/s | ~461 ms | |
#125 | Qwen3.5-9B | 9B | INT4 Q8 | 76 tok/s 48 tok/s | ~801 ms ~809 ms | |
#128 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
#135 | Gemma 4 12B | 11.95B | INT4 Q8 | 46 tok/s 29 tok/s | ~1.1s ~1.3s | |
#140 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
#143 | GPT-OSS 20B | 21B (3.6B active) | INT4 | 56 tok/s | ~474 ms | |
#144 | GLM-4V | 9B | INT4 Q8 | 57 tok/s 38 tok/s | ~806 ms ~882 ms | |
#150 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 27 tok/s 16 tok/s 9 tok/s | ~2.5s ~2.5s ~2.6s | |
#158 | Granite 4.2 8B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#159 | Gemma 4 E4B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#166 | Phi-4 | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
#167 | Gemma 3 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
#178 | OLMo 3 7B Think | 7B | INT4 Q8 | 68 tok/s 49 tok/s | ~633 ms ~638 ms | |
#179 | DeepSeek-R1 14B | 14B | INT4 Q8 | 56 tok/s 30 tok/s | ~1.2s ~1.5s | |
#183 | Ministral 3 14B | 14B | INT4 | 41 tok/s | ~1.2s | |
#186 | Gemma 3 12B | 12B | INT4 Q8 | 61 tok/s 36 tok/s | ~1.1s ~1.2s | |
#188 | Mistral-Small-2501 | 24B | INT4 | 33 tok/s | ~2.3s | |
#189 | Ministral 3 8B | 8B | INT4 Q8 | 62 tok/s 45 tok/s | ~718 ms ~725 ms | |
#191 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
#192 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
#193 | Gemma 2 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
#194 | Qwen2.5-14B | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
#200 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 110 tok/s 75 tok/s 44 tok/s | ~466 ms ~470 ms ~480 ms | |
#202 | DeepSeek-R1 8B | 8B | INT4 Q8 | 85 tok/s 54 tok/s | ~714 ms ~721 ms | |
#204 | Qwen3-4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
#205 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 106 tok/s 85 tok/s 58 tok/s | ~289 ms ~291 ms ~297 ms | |
#207 | Gemma 2 9B | 9B | INT4 Q8 | 76 tok/s 48 tok/s | ~801 ms ~809 ms | |
#208 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 225 tok/s 196 tok/s 152 tok/s | ~87 ms ~88 ms ~89 ms | |
#209 | Qwen3-14B | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
#210 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
#211 | Qwen3-8B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#213 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 93 tok/s 60 tok/s 32 tok/s | ~629 ms ~634 ms ~702 ms | |
#214 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 182 tok/s 145 tok/s 100 tok/s | ~173 ms ~175 ms ~178 ms | |
#215 | Llama 3.1 8B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#216 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 247 tok/s 235 tok/s 211 tok/s | ~35 ms ~35 ms ~35 ms | |
#218 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 238 tok/s 213 tok/s 173 tok/s | ~67 ms ~67 ms ~68 ms | |
#219 | Llama 3 8B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#221 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 130 tok/s 92 tok/s 56 tok/s | ~355 ms ~358 ms ~365 ms | |
#223 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
#224 | Ministral-8B-2410 | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~714 ms ~721 ms | |
#225 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 190 tok/s 154 tok/s 108 tok/s | ~154 ms ~156 ms ~158 ms | |
#226 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
#228 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 130 tok/s 92 tok/s 56 tok/s | ~355 ms ~358 ms ~365 ms | |
#229 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 106 tok/s 92 tok/s 66 tok/s | ~218 ms ~219 ms ~241 ms | |
#230 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 88 tok/s 57 tok/s 29 tok/s | ~654 ms ~660 ms ~789 ms | |
#231 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 213 tok/s 181 tok/s 136 tok/s | ~108 ms ~109 ms ~111 ms | |
#232 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
#233 | Phi-3-medium | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
#234 | OLMo 3 7B Instruct | 7B | INT4 Q8 | 68 tok/s 49 tok/s | ~633 ms ~638 ms | |
#236 | Qwen2-7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
#237 | Phi-3-small | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
#238 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 213 tok/s 181 tok/s 136 tok/s | ~108 ms ~109 ms ~111 ms | |
#239 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 88 tok/s 57 tok/s 29 tok/s | ~654 ms ~660 ms ~789 ms | |
#241 | Gemma 1 7B | 7B | INT4 Q8 | 68 tok/s 49 tok/s | ~633 ms ~638 ms | |
#242 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 179 tok/s 138 tok/s 92 tok/s | ~201 ms ~203 ms ~206 ms | |
#243 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 74 tok/s 55 tok/s 33 tok/s | ~547 ms ~551 ms ~609 ms | |
#244 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 74 tok/s 55 tok/s 33 tok/s | ~547 ms ~551 ms ~609 ms | |
#245 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 74 tok/s 55 tok/s 33 tok/s | ~547 ms ~551 ms ~609 ms | |
Unranked Models (Sorted by Speed) | ||||||
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 260 tok/s 245 tok/s 216 tok/s | ~37 ms ~38 ms ~38 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 246 tok/s 223 tok/s 185 tok/s | ~57 ms ~57 ms ~58 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 246 tok/s 223 tok/s 185 tok/s | ~57 ms ~57 ms ~58 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 243 tok/s 230 tok/s 204 tok/s | ~38 ms ~38 ms ~38 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 243 tok/s 230 tok/s 204 tok/s | ~38 ms ~38 ms ~38 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 238 tok/s 213 tok/s 173 tok/s | ~67 ms ~67 ms ~68 ms | |
- | PixArt-α XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 238 tok/s 212 tok/s 172 tok/s | ~67 ms ~68 ms ~69 ms | |
- | PixArt-Σ XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 238 tok/s 212 tok/s 172 tok/s | ~67 ms ~68 ms ~69 ms | |
- | Nemotron 3.5 Parse | 0.63B | INT4 Q8 FP16 | 236 tok/s 210 tok/s 169 tok/s | ~70 ms ~70 ms ~71 ms | |
- | SD 2.1 Turbo | 0.86B | INT4 Q8 FP16 | 221 tok/s 191 tok/s 147 tok/s | ~93 ms ~94 ms ~95 ms | |
- | Stable Diffusion 1.4 | 0.86B | INT4 Q8 FP16 | 221 tok/s 191 tok/s 147 tok/s | ~93 ms ~94 ms ~95 ms | |
- | Stable Diffusion 2.1 | 0.86B | INT4 Q8 FP16 | 221 tok/s 191 tok/s 147 tok/s | ~93 ms ~94 ms ~95 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 220 tok/s 186 tok/s 138 tok/s | ~108 ms ~109 ms ~111 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 213 tok/s 181 tok/s 136 tok/s | ~108 ms ~109 ms ~111 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 213 tok/s 181 tok/s 136 tok/s | ~108 ms ~109 ms ~111 ms | |
- | Kandinsky 2.2 Decoder | 1.22B | INT4 Q8 FP16 | 202 tok/s 168 tok/s 122 tok/s | ~129 ms ~130 ms ~132 ms | |
- | Wan 2.1 T2V 1.3B | 1.3B | INT4 Q8 FP16 | 199 tok/s 164 tok/s 118 tok/s | ~135 ms ~137 ms ~139 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 190 tok/s 154 tok/s 108 tok/s | ~154 ms ~156 ms ~158 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 190 tok/s 154 tok/s 108 tok/s | ~154 ms ~156 ms ~158 ms | |
- | SVD Image-to-Video XT | 1.5B | INT4 Q8 FP16 | 190 tok/s 154 tok/s 108 tok/s | ~154 ms ~156 ms ~158 ms | |
- | SVD Image-to-Video XT 1.1 | 1.5B | INT4 Q8 FP16 | 190 tok/s 154 tok/s 108 tok/s | ~154 ms ~156 ms ~158 ms | |
- | LFM3 1.7B | 1.7B | INT4 Q8 FP16 | 182 tok/s 145 tok/s 100 tok/s | ~173 ms ~175 ms ~178 ms | |
- | LFM3 1.7B Base | 1.7B | INT4 Q8 FP16 | 182 tok/s 145 tok/s 100 tok/s | ~173 ms ~175 ms ~178 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 182 tok/s 145 tok/s 100 tok/s | ~173 ms ~175 ms ~178 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 179 tok/s 141 tok/s 96 tok/s | ~182 ms ~184 ms ~187 ms | |
- | Moondream 3.1 | 9B (2B active) | INT4 Q8 | 172 tok/s 134 tok/s | ~226 ms ~227 ms | |
- | CogVideoX 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | LFM3 Audio 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | LFM3 Omni 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | LFM3 VL 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | Liquid D1 2B | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | LTX-Video 0.9.1 | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | Stable Diffusion 3 Medium | 2B | INT4 Q8 FP16 | 171 tok/s 134 tok/s 90 tok/s | ~201 ms ~203 ms ~207 ms | |
- | Kandinsky 3.0 | 2.1B | INT4 Q8 FP16 | 168 tok/s 130 tok/s 87 tok/s | ~210 ms ~212 ms ~215 ms | |
- | LTX-2 | 2.2B | INT4 Q8 FP16 | 166 tok/s 127 tok/s 84 tok/s | ~218 ms ~220 ms ~224 ms | |
- | SDXL 1.0 Refiner | 2.31B | INT4 Q8 FP16 | 163 tok/s 124 tok/s 81 tok/s | ~228 ms ~230 ms ~234 ms | |
- | Stable Diffusion 3.5 Medium | 2.5B | INT4 Q8 FP16 | 157 tok/s 119 tok/s 77 tok/s | ~244 ms ~246 ms ~250 ms | |
- | SDXL 1.0 Base | 2.56B | INT4 Q8 FP16 | 156 tok/s 117 tok/s 76 tok/s | ~249 ms ~251 ms ~256 ms | |
- | SDXL Turbo | 2.56B | INT4 Q8 FP16 | 156 tok/s 117 tok/s 76 tok/s | ~249 ms ~251 ms ~256 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 152 tok/s 110 tok/s 68 tok/s | ~286 ms ~289 ms ~294 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 150 tok/s 131 tok/s 100 tok/s | ~137 ms ~138 ms ~140 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 150 tok/s 131 tok/s 100 tok/s | ~137 ms ~138 ms ~140 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 150 tok/s 131 tok/s 100 tok/s | ~137 ms ~138 ms ~140 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 147 tok/s 108 tok/s 68 tok/s | ~286 ms ~289 ms ~294 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 | 147 tok/s 108 tok/s | ~280 ms ~305 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 146 tok/s 107 tok/s 67 tok/s | ~286 ms ~289 ms ~294 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 | 131 tok/s | ~343 ms | |
- | FLUX.2 Klein 4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
- | Mage-VL Cowork | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 126 tok/s 89 tok/s 54 tok/s | ~372 ms ~375 ms ~383 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 | 125 tok/s | ~377 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 | 125 tok/s | ~377 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 122 tok/s 104 tok/s 75 tok/s | ~204 ms ~205 ms ~209 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 | 115 tok/s | ~428 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 | 115 tok/s | ~428 ms | |
- | CogVideoX 1.5 5B | 5B | INT4 Q8 FP16 | 112 tok/s 76 tok/s 45 tok/s | ~458 ms ~462 ms ~471 ms | |
- | CogVideoX 5B | 5B | INT4 Q8 FP16 | 112 tok/s 76 tok/s 45 tok/s | ~458 ms ~462 ms ~471 ms | |
- | Wan 2.2 TI2V 5B | 5B | INT4 Q8 FP16 | 112 tok/s 76 tok/s 45 tok/s | ~458 ms ~462 ms ~471 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 110 tok/s 90 tok/s 62 tok/s | ~263 ms ~265 ms ~270 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 106 tok/s 85 tok/s 58 tok/s | ~289 ms ~291 ms ~297 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 106 tok/s 85 tok/s 58 tok/s | ~289 ms ~291 ms ~297 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 32 tok/s | ~628 ms ~634 ms ~702 ms | |
- | AuraFlow v0.2 | 6.8B | INT4 Q8 FP16 | 92 tok/s 61 tok/s 33 tok/s | ~612 ms ~617 ms ~683 ms | |
- | AuraFlow v0.3 | 6.8B | INT4 Q8 FP16 | 92 tok/s 61 tok/s 33 tok/s | ~612 ms ~617 ms ~683 ms | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 30 tok/s | ~629 ms ~635 ms ~758 ms | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 88 tok/s 57 tok/s 29 tok/s | ~654 ms ~660 ms ~789 ms | |
- | Youtu Parsing Omni | 8B | INT4 Q8 | 85 tok/s 54 tok/s | ~714 ms ~721 ms | |
- | Stable Diffusion 3.5 Large | 8.1B | INT4 Q8 | 82 tok/s 53 tok/s | ~723 ms ~729 ms | |
- | Stable Diffusion 3.5 Large Turbo | 8.1B | INT4 Q8 | 82 tok/s 53 tok/s | ~723 ms ~729 ms | |
- | FLUX.2 Klein 9B | 9B | INT4 Q8 | 76 tok/s 48 tok/s | ~801 ms ~809 ms | |
- | LensVLM 9B | 9B | INT4 Q8 | 76 tok/s 48 tok/s | ~801 ms ~809 ms | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 | 76 tok/s 48 tok/s | ~801 ms ~809 ms | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 74 tok/s 55 tok/s 33 tok/s | ~547 ms ~551 ms ~609 ms | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 74 tok/s 55 tok/s 33 tok/s | ~547 ms ~551 ms ~609 ms | |
- | Falcon3-10B | 10B | INT4 Q8 | 71 tok/s 44 tok/s | ~887 ms ~895 ms | |
- | Mochi 1 Preview | 10B | INT4 Q8 | 71 tok/s 44 tok/s | ~887 ms ~895 ms | |
- | Falcon2-11B | 11B | INT4 Q8 | 68 tok/s 42 tok/s | ~972 ms ~982 ms | |
- | Hunyuan Lite | 7B | INT4 Q8 | 68 tok/s 49 tok/s | ~633 ms ~638 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 | 68 tok/s 49 tok/s | ~633 ms ~638 ms | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 | 67 tok/s | ~369 ms | |
- | SEA-LION-7B | 7.1B | INT4 Q8 | 67 tok/s 49 tok/s | ~641 ms ~647 ms | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 | 67 tok/s 49 tok/s | ~641 ms ~647 ms | |
- | FLUX.1 Canny Dev | 11.9B | INT4 Q8 | 62 tok/s 36 tok/s | ~1.1s ~1.2s | |
- | FLUX.1 Depth Dev | 11.9B | INT4 Q8 | 62 tok/s 36 tok/s | ~1.1s ~1.2s | |
- | FLUX.1 Dev | 11.9B | INT4 Q8 | 62 tok/s 36 tok/s | ~1.1s ~1.2s | |
- | FLUX.1 Fill Dev | 11.9B | INT4 Q8 | 62 tok/s 36 tok/s | ~1.1s ~1.2s | |
- | FLUX.1 Schnell | 11.9B | INT4 Q8 | 62 tok/s 36 tok/s | ~1.1s ~1.2s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 | 62 tok/s 45 tok/s | ~718 ms ~725 ms | |
- | Typhoon-2-8B | 8B | INT4 Q8 | 62 tok/s 45 tok/s | ~718 ms ~725 ms | |
- | HunyuanVideo 1.5 720P T2V | 13B | INT4 Q8 | 58 tok/s 34 tok/s | ~1.1s ~1.3s | |
- | GLM-4-9B | 9B | INT4 Q8 | 57 tok/s 38 tok/s | ~806 ms ~882 ms | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 | 57 tok/s 38 tok/s | ~806 ms ~882 ms | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 | 57 tok/s 38 tok/s | ~806 ms ~882 ms | |
- | Yi-9B | 9B | INT4 Q8 | 57 tok/s 38 tok/s | ~806 ms ~882 ms | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 | 56 tok/s 38 tok/s | ~823 ms ~900 ms | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 | 56 tok/s 38 tok/s | ~823 ms ~900 ms | |
- | Wan 2.1 I2V 14B 720P | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
- | Wan 2.1 T2V 14B | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
- | Wan 2.2 T2V A14B | 14B | INT4 Q8 | 54 tok/s 30 tok/s | ~1.2s ~1.5s | |
- | FLUX.2 Dev | 15B | INT4 Q8 | 52 tok/s 28 tok/s | ~1.3s ~1.6s | |
- | GreenMind-14B-R1 | 14B | INT4 | 41 tok/s | ~1.2s | |
- | Arete 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
- | AREX-2 | 27B | INT4 | 28 tok/s | ~2.7s | |
- | PPLX Decider v1 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
- | PPLX Decider v1.1 27B | 27B | INT4 | 28 tok/s | ~2.7s | |
Inference Capacity
Maximum model parameter class runnable by weight precision.
Context:
4-bit
14B Models
8-bit
8B Models
16-bit
3B Models
Fine-Tuning Capacity
Hardware capacity for local training and adapter fine-tuning.
Context Length:
QLoRA
14B Models
LoRA
8B Models
Full Parameter Training
1B Models
Workload Recommendations
Guidance for optimal precision and operational ceilings on AMD RX 9070 (16GB).
Inference Sweet Spot
Optimal for 8B models across extended context (32k+). 14B models fit comfortably at Q4 precision with standard context.
Bandwidth & Speed Profile
With 640 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 142 tok/s on an 8B Q4 model and 17 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning