AMD RX 7800 XT (16GB)
The AMD RX 7800 XT (16GB) provides 16 GB of memory with 624 GB/s bandwidth, supporting local inference via ROCm and Vulkan acceleration.
Memory
16 GB
Usable Memory Ceiling
16 GB
Bandwidth
624 GB/s
Max Inference Class
14B Models (Q4)
TDP
263W
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 7800 XT (16GB)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on AMD RX 7800 XT (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 | ~5.3s | |
#93 | Qwen3.5-27B | 27B | INT4 | 28 tok/s | ~5.3s | |
#111 | MiMo V2 Flash | 15B (309B active) | INT4 | 3 tok/s | ~53.8s | |
#112 | Phi-4 Reasoning Plus | 14B | INT4 | 40 tok/s | ~2.4s | |
#117 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 | 109 tok/s | ~886 ms | |
#125 | Qwen3.5-9B | 9B | INT4 Q8 | 75 tok/s 47 tok/s | ~1.5s ~1.6s | |
#128 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
#135 | Gemma 4 12B | 11.95B | INT4 Q8 | 45 tok/s 28 tok/s | ~2.0s ~2.4s | |
#140 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
#143 | GPT-OSS 20B | 21B (3.6B active) | INT4 | 55 tok/s | ~905 ms | |
#144 | GLM-4V | 9B | INT4 Q8 | 56 tok/s 38 tok/s | ~1.6s ~1.7s | |
#150 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 27 tok/s 16 tok/s 8 tok/s | ~4.9s ~4.9s ~4.9s | |
#158 | Granite 4.2 8B | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#159 | Gemma 4 E4B | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#166 | Phi-4 | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
#167 | Gemma 3 27B | 27B | INT4 | 28 tok/s | ~5.3s | |
#178 | OLMo 3 7B Think | 7B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
#179 | DeepSeek-R1 14B | 14B | INT4 Q8 | 55 tok/s 30 tok/s | ~2.4s ~2.8s | |
#183 | Ministral 3 14B | 14B | INT4 | 40 tok/s | ~2.4s | |
#186 | Gemma 3 12B | 12B | INT4 Q8 | 60 tok/s 35 tok/s | ~2.1s ~2.2s | |
#188 | Mistral-Small-2501 | 24B | INT4 | 32 tok/s | ~4.4s | |
#189 | Ministral 3 8B | 8B | INT4 Q8 | 61 tok/s 44 tok/s | ~1.4s ~1.4s | |
#191 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
#192 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
#193 | Gemma 2 27B | 27B | INT4 | 28 tok/s | ~5.3s | |
#194 | Qwen2.5-14B | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
#200 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 109 tok/s 74 tok/s 43 tok/s | ~897 ms ~901 ms ~911 ms | |
#202 | DeepSeek-R1 8B | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~1.4s ~1.4s | |
#204 | Qwen3-4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
#205 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 104 tok/s 84 tok/s 57 tok/s | ~550 ms ~552 ms ~558 ms | |
#207 | Gemma 2 9B | 9B | INT4 Q8 | 75 tok/s 47 tok/s | ~1.5s ~1.6s | |
#208 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 224 tok/s 194 tok/s 150 tok/s | ~160 ms ~161 ms ~162 ms | |
#209 | Qwen3-14B | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
#210 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
#211 | Qwen3-8B | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#213 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 31 tok/s | ~1.2s ~1.2s ~1.3s | |
#214 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 180 tok/s 143 tok/s 98 tok/s | ~328 ms ~329 ms ~333 ms | |
#215 | Llama 3.1 8B | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#216 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 247 tok/s 234 tok/s 209 tok/s | ~58 ms ~59 ms ~59 ms | |
#218 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 237 tok/s 212 tok/s 171 tok/s | ~120 ms ~121 ms ~122 ms | |
#219 | Llama 3 8B | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#221 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 128 tok/s 91 tok/s 55 tok/s | ~680 ms ~684 ms ~691 ms | |
#223 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
#224 | Ministral-8B-2410 | 8B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
#225 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 188 tok/s 152 tok/s 106 tok/s | ~291 ms ~292 ms ~295 ms | |
#226 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
#228 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 128 tok/s 91 tok/s 55 tok/s | ~680 ms ~684 ms ~691 ms | |
#229 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 104 tok/s 91 tok/s 64 tok/s | ~411 ms ~413 ms ~451 ms | |
#230 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 87 tok/s 56 tok/s 29 tok/s | ~1.3s ~1.3s ~1.5s | |
#231 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 211 tok/s 179 tok/s 134 tok/s | ~201 ms ~202 ms ~204 ms | |
#232 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
#233 | Phi-3-medium | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
#234 | OLMo 3 7B Instruct | 7B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
#236 | Qwen2-7B | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
#237 | Phi-3-small | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
#238 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 211 tok/s 179 tok/s 134 tok/s | ~201 ms ~202 ms ~204 ms | |
#239 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 87 tok/s 56 tok/s 29 tok/s | ~1.3s ~1.3s ~1.5s | |
#241 | Gemma 1 7B | 7B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
#242 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 178 tok/s 136 tok/s 90 tok/s | ~382 ms ~384 ms ~387 ms | |
#243 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 73 tok/s 54 tok/s 32 tok/s | ~1.1s ~1.1s ~1.2s | |
#244 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 73 tok/s 54 tok/s 32 tok/s | ~1.1s ~1.1s ~1.2s | |
#245 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 73 tok/s 54 tok/s 32 tok/s | ~1.1s ~1.1s ~1.2s | |
Unranked Models (Sorted by Speed) | ||||||
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 260 tok/s 244 tok/s 214 tok/s | ~64 ms ~64 ms ~65 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 245 tok/s 222 tok/s 183 tok/s | ~101 ms ~101 ms ~102 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 245 tok/s 222 tok/s 183 tok/s | ~101 ms ~101 ms ~102 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 242 tok/s 229 tok/s 203 tok/s | ~64 ms ~64 ms ~65 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 242 tok/s 229 tok/s 203 tok/s | ~64 ms ~64 ms ~65 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 237 tok/s 212 tok/s 171 tok/s | ~120 ms ~121 ms ~122 ms | |
- | PixArt-α XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 237 tok/s 211 tok/s 170 tok/s | ~122 ms ~122 ms ~124 ms | |
- | PixArt-Σ XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 237 tok/s 211 tok/s 170 tok/s | ~122 ms ~122 ms ~124 ms | |
- | Nemotron 3.5 Parse | 0.63B | INT4 Q8 FP16 | 235 tok/s 209 tok/s 167 tok/s | ~127 ms ~127 ms ~128 ms | |
- | SD 2.1 Turbo | 0.86B | INT4 Q8 FP16 | 220 tok/s 190 tok/s 145 tok/s | ~172 ms ~173 ms ~174 ms | |
- | Stable Diffusion 1.4 | 0.86B | INT4 Q8 FP16 | 220 tok/s 190 tok/s 145 tok/s | ~172 ms ~173 ms ~174 ms | |
- | Stable Diffusion 2.1 | 0.86B | INT4 Q8 FP16 | 220 tok/s 190 tok/s 145 tok/s | ~172 ms ~173 ms ~174 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 219 tok/s 184 tok/s 137 tok/s | ~201 ms ~202 ms ~204 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 211 tok/s 179 tok/s 134 tok/s | ~201 ms ~202 ms ~204 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 211 tok/s 179 tok/s 134 tok/s | ~201 ms ~202 ms ~204 ms | |
- | Kandinsky 2.2 Decoder | 1.22B | INT4 Q8 FP16 | 201 tok/s 166 tok/s 120 tok/s | ~241 ms ~242 ms ~244 ms | |
- | Wan 2.1 T2V 1.3B | 1.3B | INT4 Q8 FP16 | 197 tok/s 162 tok/s 116 tok/s | ~254 ms ~256 ms ~258 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 188 tok/s 152 tok/s 106 tok/s | ~291 ms ~292 ms ~295 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 188 tok/s 152 tok/s 106 tok/s | ~291 ms ~292 ms ~295 ms | |
- | SVD Image-to-Video XT | 1.5B | INT4 Q8 FP16 | 188 tok/s 152 tok/s 106 tok/s | ~291 ms ~292 ms ~295 ms | |
- | SVD Image-to-Video XT 1.1 | 1.5B | INT4 Q8 FP16 | 188 tok/s 152 tok/s 106 tok/s | ~291 ms ~292 ms ~295 ms | |
- | LFM3 1.7B | 1.7B | INT4 Q8 FP16 | 180 tok/s 143 tok/s 98 tok/s | ~328 ms ~329 ms ~333 ms | |
- | LFM3 1.7B Base | 1.7B | INT4 Q8 FP16 | 180 tok/s 143 tok/s 98 tok/s | ~328 ms ~329 ms ~333 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 180 tok/s 143 tok/s 98 tok/s | ~328 ms ~329 ms ~333 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 177 tok/s 139 tok/s 95 tok/s | ~345 ms ~347 ms ~350 ms | |
- | CogVideoX 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | LFM3 Audio 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | LFM3 Omni 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | LFM3 VL 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | Liquid D1 2B | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | LTX-Video 0.9.1 | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | Moondream 3.1 | 9B (2B active) | INT4 Q8 | 170 tok/s 132 tok/s | ~430 ms ~431 ms | |
- | Stable Diffusion 3 Medium | 2B | INT4 Q8 FP16 | 170 tok/s 132 tok/s 88 tok/s | ~382 ms ~384 ms ~388 ms | |
- | Kandinsky 3.0 | 2.1B | INT4 Q8 FP16 | 167 tok/s 129 tok/s 85 tok/s | ~399 ms ~401 ms ~404 ms | |
- | LTX-2 | 2.2B | INT4 Q8 FP16 | 164 tok/s 125 tok/s 83 tok/s | ~415 ms ~417 ms ~421 ms | |
- | SDXL 1.0 Refiner | 2.31B | INT4 Q8 FP16 | 161 tok/s 122 tok/s 80 tok/s | ~433 ms ~435 ms ~440 ms | |
- | Stable Diffusion 3.5 Medium | 2.5B | INT4 Q8 FP16 | 156 tok/s 117 tok/s 75 tok/s | ~465 ms ~467 ms ~472 ms | |
- | SDXL 1.0 Base | 2.56B | INT4 Q8 FP16 | 154 tok/s 115 tok/s 74 tok/s | ~475 ms ~477 ms ~482 ms | |
- | SDXL Turbo | 2.56B | INT4 Q8 FP16 | 154 tok/s 115 tok/s 74 tok/s | ~475 ms ~477 ms ~482 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 150 tok/s 108 tok/s 67 tok/s | ~547 ms ~550 ms ~555 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 148 tok/s 130 tok/s 98 tok/s | ~256 ms ~257 ms ~260 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 148 tok/s 130 tok/s 98 tok/s | ~256 ms ~257 ms ~260 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 148 tok/s 130 tok/s 98 tok/s | ~256 ms ~257 ms ~260 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 | 146 tok/s 106 tok/s | ~535 ms ~582 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 145 tok/s 106 tok/s 66 tok/s | ~547 ms ~550 ms ~555 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 144 tok/s 105 tok/s 66 tok/s | ~547 ms ~550 ms ~556 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 | 130 tok/s | ~657 ms | |
- | FLUX.2 Klein 4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
- | Mage-VL Cowork | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 125 tok/s 87 tok/s 53 tok/s | ~714 ms ~717 ms ~725 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 | 123 tok/s | ~725 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 | 123 tok/s | ~725 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 120 tok/s 102 tok/s 74 tok/s | ~385 ms ~386 ms ~390 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 | 113 tok/s | ~824 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 | 113 tok/s | ~824 ms | |
- | CogVideoX 1.5 5B | 5B | INT4 Q8 FP16 | 110 tok/s 75 tok/s 44 tok/s | ~880 ms ~884 ms ~894 ms | |
- | CogVideoX 5B | 5B | INT4 Q8 FP16 | 110 tok/s 75 tok/s 44 tok/s | ~880 ms ~884 ms ~894 ms | |
- | Wan 2.2 TI2V 5B | 5B | INT4 Q8 FP16 | 110 tok/s 75 tok/s 44 tok/s | ~880 ms ~884 ms ~894 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 109 tok/s 89 tok/s 61 tok/s | ~500 ms ~502 ms ~508 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 104 tok/s 84 tok/s 57 tok/s | ~550 ms ~552 ms ~558 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 104 tok/s 84 tok/s 57 tok/s | ~550 ms ~552 ms ~558 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 93 tok/s 59 tok/s 32 tok/s | ~1.2s ~1.2s ~1.3s | |
- | AuraFlow v0.2 | 6.8B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 32 tok/s | ~1.2s ~1.2s ~1.3s | |
- | AuraFlow v0.3 | 6.8B | INT4 Q8 FP16 | 91 tok/s 59 tok/s 32 tok/s | ~1.2s ~1.2s ~1.3s | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 89 tok/s 58 tok/s 30 tok/s | ~1.2s ~1.2s ~1.4s | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 87 tok/s 56 tok/s 29 tok/s | ~1.3s ~1.3s ~1.5s | |
- | Youtu Parsing Omni | 8B | INT4 Q8 | 83 tok/s 53 tok/s | ~1.4s ~1.4s | |
- | Stable Diffusion 3.5 Large | 8.1B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
- | Stable Diffusion 3.5 Large Turbo | 8.1B | INT4 Q8 | 81 tok/s 52 tok/s | ~1.4s ~1.4s | |
- | FLUX.2 Klein 9B | 9B | INT4 Q8 | 75 tok/s 47 tok/s | ~1.5s ~1.6s | |
- | LensVLM 9B | 9B | INT4 Q8 | 75 tok/s 47 tok/s | ~1.5s ~1.6s | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 | 75 tok/s 47 tok/s | ~1.5s ~1.6s | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 73 tok/s 54 tok/s 32 tok/s | ~1.1s ~1.1s ~1.2s | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 73 tok/s 54 tok/s 32 tok/s | ~1.1s ~1.1s ~1.2s | |
- | Falcon3-10B | 10B | INT4 Q8 | 69 tok/s 43 tok/s | ~1.7s ~1.7s | |
- | Mochi 1 Preview | 10B | INT4 Q8 | 69 tok/s 43 tok/s | ~1.7s ~1.7s | |
- | Falcon2-11B | 11B | INT4 Q8 | 67 tok/s 41 tok/s | ~1.9s ~1.9s | |
- | Hunyuan Lite | 7B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 | 66 tok/s | ~703 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
- | SEA-LION-7B | 7.1B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 | 66 tok/s 48 tok/s | ~1.2s ~1.2s | |
- | FLUX.1 Canny Dev | 11.9B | INT4 Q8 | 61 tok/s 36 tok/s | ~2.0s ~2.2s | |
- | FLUX.1 Depth Dev | 11.9B | INT4 Q8 | 61 tok/s 36 tok/s | ~2.0s ~2.2s | |
- | FLUX.1 Dev | 11.9B | INT4 Q8 | 61 tok/s 36 tok/s | ~2.0s ~2.2s | |
- | FLUX.1 Fill Dev | 11.9B | INT4 Q8 | 61 tok/s 36 tok/s | ~2.0s ~2.2s | |
- | FLUX.1 Schnell | 11.9B | INT4 Q8 | 61 tok/s 36 tok/s | ~2.0s ~2.2s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 | 61 tok/s 44 tok/s | ~1.4s ~1.4s | |
- | Typhoon-2-8B | 8B | INT4 Q8 | 61 tok/s 44 tok/s | ~1.4s ~1.4s | |
- | HunyuanVideo 1.5 720P T2V | 13B | INT4 Q8 | 57 tok/s 33 tok/s | ~2.2s ~2.4s | |
- | GLM-4-9B | 9B | INT4 Q8 | 56 tok/s 38 tok/s | ~1.6s ~1.7s | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 | 56 tok/s 38 tok/s | ~1.6s ~1.7s | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 | 56 tok/s 38 tok/s | ~1.6s ~1.7s | |
- | Yi-9B | 9B | INT4 Q8 | 56 tok/s 38 tok/s | ~1.6s ~1.7s | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 | 55 tok/s 37 tok/s | ~1.6s ~1.7s | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 | 55 tok/s 37 tok/s | ~1.6s ~1.7s | |
- | Wan 2.1 I2V 14B 720P | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
- | Wan 2.1 T2V 14B | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
- | Wan 2.2 T2V A14B | 14B | INT4 Q8 | 53 tok/s 29 tok/s | ~2.4s ~2.8s | |
- | FLUX.2 Dev | 15B | INT4 Q8 | 51 tok/s 28 tok/s | ~2.5s ~3.0s | |
- | GreenMind-14B-R1 | 14B | INT4 | 40 tok/s | ~2.4s | |
- | Arete 27B | 27B | INT4 | 28 tok/s | ~5.3s | |
- | AREX-2 | 27B | INT4 | 28 tok/s | ~5.3s | |
- | PPLX Decider v1 27B | 27B | INT4 | 28 tok/s | ~5.3s | |
- | PPLX Decider v1.1 27B | 27B | INT4 | 28 tok/s | ~5.3s | |
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 7800 XT (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 624 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 139 tok/s on an 8B Q4 model and 16 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning