NVIDIA GB10 Grace Blackwell (128GB) - DGX Spark
The NVIDIA GB10 Grace Blackwell (128GB) - DGX Spark features 128 GB of dedicated VRAM and 273 GB/s memory bandwidth, supporting CUDA and TensorRT-LLM runtimes for full-precision and quantized inference.
Memory
128 GB
Usable Memory Ceiling
128 GB
Bandwidth
273.2 GB/s
Max Inference Class
70B+ Models (Q4)
TDP
140W
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 NVIDIA GB10 Grace Blackwell (128GB) - DGX Spark
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on NVIDIA GB10 Grace Blackwell (128GB) - DGX Spark.
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#36 | Qwen 3.8 27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
#42 | Qwen 3.8 Flash | 176B (6B active) | INT4 | 59 tok/s | ~440 ms | |
#69 | Sarvam-105B | 106B (10.3B active) | INT4 Q8 | 35 tok/s 19 tok/s | ~1.0s ~1.1s | |
#80 | K2 Horizon MoVA 36B A4B | 36B (4B active) | INT4 Q8 FP16 | 64 tok/s 43 tok/s 25 tok/s | ~400 ms ~408 ms ~425 ms | |
#83 | Ling 3.0 Flash VL | 124B (5.5B active) | INT4 Q8 | 57 tok/s 32 tok/s | ~731 ms ~869 ms | |
#86 | Qwen3.5-27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
#88 | Qwen3.5-122B-A10B | 122B (10B active) | INT4 Q8 | 27 tok/s 16 tok/s | ~1.1s ~1.3s | |
#93 | Sarvam-30B | 32B (2.4B active) | INT4 Q8 FP16 | 85 tok/s 62 tok/s 38 tok/s | ~281 ms ~285 ms ~296 ms | |
#99 | Muse Glimmer 30B | 30B | INT4 Q8 FP16 | 13 tok/s 7 tok/s 4 tok/s | ~2.2s ~2.3s ~2.4s | |
#102 | Phi-4 Reasoning Plus | 14B | INT4 Q8 FP16 | 19 tok/s 13 tok/s 7 tok/s | ~1.1s ~1.1s ~1.2s | |
#103 | Qwen3.6 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 76 tok/s 53 tok/s 31 tok/s | ~327 ms ~332 ms ~345 ms | |
#105 | MiMo V2 Flash | 15B (309B active) | INT4 Q8 FP16 | 1 tok/s 1 tok/s 1 tok/s | ~22.1s ~22.7s ~24.0s | |
#110 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 Q8 FP16 | 69 tok/s 46 tok/s 26 tok/s | ~367 ms ~374 ms ~390 ms | |
#113 | GLM-4.5-Air | 106B (12B active) | INT4 Q8 | 9 tok/s 7 tok/s | ~1.2s ~1.5s | |
#117 | Nemotron 3.5 Bolt 74B A9B | 74B | INT4 Q8 | 5 tok/s 3 tok/s | ~5.5s ~5.6s | |
#120 | Qwen3.5-35B-A3B | 35B (3B active) | INT4 Q8 FP16 | 76 tok/s 53 tok/s 31 tok/s | ~327 ms ~332 ms ~345 ms | |
#121 | Qwen3.5-9B | 9B | INT4 Q8 FP16 | 40 tok/s 23 tok/s 12 tok/s | ~693 ms ~710 ms ~749 ms | |
#122 | Ling 3.0 Flash | 124B (5.1B active) | INT4 Q8 | 60 tok/s 34 tok/s | ~702 ms ~835 ms | |
#123 | Step 3.5 Flash | 196.81B (11B active) | INT4 | 23 tok/s | ~1.3s | |
#127 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
#130 | Gemma 4 12B | 11.95B | INT4 Q8 FP16 | 22 tok/s 15 tok/s 8 tok/s | ~926 ms ~948 ms ~1.0s | |
#131 | Qwen3 Next 80B A3B | 80B (79B active) | INT4 Q8 | 4 tok/s 3 tok/s | ~5.9s ~6.0s | |
#133 | Qwen3-235B-A22B | 235B (22B active) | INT4 | 13 tok/s | ~2.4s | |
#135 | GPT-OSS 120B | 117B (5.1B active) | INT4 | 10 tok/s | ~778 ms | |
#136 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 77 tok/s 48 tok/s 26 tok/s | ~321 ms ~329 ms ~346 ms | |
#139 | GPT-OSS 20B | 21B (3.6B active) | INT4 Q8 FP16 | 31 tok/s 27 tok/s 19 tok/s | ~362 ms ~367 ms ~383 ms | |
#140 | GLM-4V | 9B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~703 ms ~720 ms ~759 ms | |
#145 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 12 tok/s 7 tok/s 4 tok/s | ~2.2s ~2.2s ~2.4s | |
#147 | Mistral Medium 3.5 128B | 128B | INT4 | 3 tok/s | ~9.5s | |
#147 | Nemotron 3.5 Super VL | 120B | INT4 Q8 | 3 tok/s 2 tok/s | ~8.9s ~10.7s | |
#150 | Qwen3-30B-A3B | 30B (3B active) | INT4 Q8 FP16 | 77 tok/s 54 tok/s 32 tok/s | ~321 ms ~326 ms ~339 ms | |
#152 | Granite 4.2 8B | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#157 | Gemma 4 E4B | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#158 | Mistral Large 3 | 41B (675B active) | INT4 Q8 FP16 | 1 tok/s 1 tok/s 1 tok/s | ~48.1s ~49.5s ~52.4s | |
#159 | Qwen2.5-72B | 72B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.3s ~5.5s | |
#161 | GLM-4 | 32B | INT4 Q8 FP16 | 10 tok/s 6 tok/s 3 tok/s | ~2.4s ~2.5s ~2.6s | |
#162 | DeepSeek-R1 70B | 70B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.2s ~5.3s | |
#163 | Gemma 3 27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
#164 | Phi-4 | 14B | INT4 Q8 FP16 | 27 tok/s 15 tok/s 8 tok/s | ~1.1s ~1.1s ~1.2s | |
#165 | Llama 3.3 70B | 70B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.2s ~5.3s | |
#166 | Qwen3-32B | 32B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 4 tok/s | ~2.4s ~2.4s ~2.6s | |
#167 | Hunyuan Standard | 52B (389B active) | INT4 Q8 FP16 | 1 tok/s 1 tok/s 1 tok/s | ~27.8s ~28.6s ~32.8s | |
#168 | DeepSeek-R1 32B | 32B | INT4 Q8 FP16 | 13 tok/s 7 tok/s 4 tok/s | ~2.4s ~2.4s ~2.6s | |
#169 | Gemma 4 31B | 30.7B | INT4 Q8 FP16 | 13 tok/s 7 tok/s 4 tok/s | ~2.3s ~2.4s ~2.5s | |
#171 | Magistral Small | 24B | INT4 Q8 FP16 | 12 tok/s 8 tok/s 4 tok/s | ~1.8s ~1.9s ~2.0s | |
#172 | Qwen2.5-32B | 32B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 4 tok/s | ~2.4s ~2.4s ~2.6s | |
#174 | OLMo 3.1 32B Instruct | 32B | INT4 Q8 FP16 | 10 tok/s 6 tok/s 3 tok/s | ~2.4s ~2.5s ~2.6s | |
#175 | Llama 3.1 70B | 70B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.2s ~5.3s | |
#176 | DeepSeek-R1 14B | 14B | INT4 Q8 FP16 | 27 tok/s 16 tok/s 8 tok/s | ~1.1s ~1.1s ~1.2s | |
#177 | Ministral 3 14B | 14B | INT4 Q8 FP16 | 19 tok/s 13 tok/s 7 tok/s | ~1.1s ~1.1s ~1.2s | |
#178 | Mistral-Large-2407 | 123B | INT4 Q8 | 3 tok/s 2 tok/s | ~9.1s ~10.9s | |
#179 | Llama 4 Scout | 109B (17B active) | INT4 Q8 | 19 tok/s 11 tok/s | ~1.5s ~1.7s | |
#180 | Gemma 3 12B | 12B | INT4 Q8 FP16 | 31 tok/s 18 tok/s 9 tok/s | ~917 ms ~940 ms ~992 ms | |
#182 | Command A | 111B | INT4 Q8 | 3 tok/s 2 tok/s | ~8.3s ~9.9s | |
#183 | Qwen2-72B | 72B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.3s ~5.5s | |
#184 | Mistral-Small-2501 | 24B | INT4 Q8 FP16 | 16 tok/s 9 tok/s 5 tok/s | ~1.8s ~1.8s ~2.0s | |
#185 | Ministral 3 8B | 8B | INT4 Q8 FP16 | 31 tok/s 21 tok/s 12 tok/s | ~627 ms ~642 ms ~676 ms | |
#186 | Gemma 2 27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
#187 | OLMo 3 7B Think | 7B | INT4 Q8 FP16 | 34 tok/s 24 tok/s 14 tok/s | ~552 ms ~565 ms ~595 ms | |
#188 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 122 tok/s 83 tok/s 48 tok/s | ~174 ms ~177 ms ~186 ms | |
#189 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 122 tok/s 83 tok/s 48 tok/s | ~174 ms ~177 ms ~186 ms | |
#190 | Qwen2.5-14B | 14B | INT4 Q8 FP16 | 27 tok/s 15 tok/s 8 tok/s | ~1.1s ~1.1s ~1.2s | |
#191 | OLMo 3.1 32B Think | 32B | INT4 Q8 FP16 | 10 tok/s 6 tok/s 3 tok/s | ~2.4s ~2.5s ~2.6s | |
#192 | Llama 3 70B | 70B | INT4 Q8 | 6 tok/s 3 tok/s | ~5.2s ~5.3s | |
#193 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 64 tok/s 39 tok/s 21 tok/s | ~403 ms ~413 ms ~435 ms | |
#194 | Command R Plus | 104B | INT4 Q8 | 3 tok/s 2 tok/s | ~7.8s ~8.6s | |
#195 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 60 tok/s 46 tok/s 29 tok/s | ~253 ms ~258 ms ~271 ms | |
#196 | DeepSeek-R1 8B | 8B | INT4 Q8 FP16 | 45 tok/s 26 tok/s 14 tok/s | ~617 ms ~632 ms ~667 ms | |
#198 | Gemma 2 9B | 9B | INT4 Q8 FP16 | 40 tok/s 23 tok/s 12 tok/s | ~693 ms ~710 ms ~749 ms | |
#199 | Nemotron 3.5 Lightning 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 77 tok/s 54 tok/s 32 tok/s | ~321 ms ~326 ms ~339 ms | |
#200 | Kimi Linear 48B A3B Instruct | 48B (3B active) | INT4 Q8 FP16 | 23 tok/s 22 tok/s 16 tok/s | ~375 ms ~378 ms ~422 ms | |
#201 | Qwen3-4B | 4B | INT4 Q8 FP16 | 77 tok/s 48 tok/s 26 tok/s | ~321 ms ~329 ms ~346 ms | |
#202 | Qwen3-14B | 14B | INT4 Q8 FP16 | 27 tok/s 15 tok/s 8 tok/s | ~1.1s ~1.1s ~1.2s | |
#203 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 77 tok/s 48 tok/s 26 tok/s | ~321 ms ~329 ms ~346 ms | |
#204 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 51 tok/s 30 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
#205 | OLMo 3 32B Base | 32B | INT4 Q8 FP16 | 10 tok/s 6 tok/s 3 tok/s | ~2.4s ~2.5s ~2.6s | |
#206 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 204 tok/s 155 tok/s 101 tok/s | ~75 ms ~76 ms ~80 ms | |
#207 | Llama 3.1 8B | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#207 | Qwen3-8B | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#209 | Mixtral-8x22B-v0.1 | 176B (22B active) | INT4 | 15 tok/s | ~1.9s | |
#210 | Nemotron 3.5 Lightning | 30B (3B active) | INT4 Q8 FP16 | 28 tok/s 25 tok/s 19 tok/s | ~344 ms ~347 ms ~360 ms | |
#211 | Command R | 35B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~2.6s ~2.7s ~2.8s | |
#212 | Llama 3 8B | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#213 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 136 tok/s 94 tok/s 56 tok/s | ~149 ms ~153 ms ~160 ms | |
#214 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 252 tok/s 225 tok/s 178 tok/s | ~30 ms ~30 ms ~31 ms | |
#216 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 231 tok/s 182 tok/s 124 tok/s | ~57 ms ~58 ms ~61 ms | |
#217 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 80 tok/s 50 tok/s 28 tok/s | ~306 ms ~314 ms ~330 ms | |
#218 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
#219 | Mixtral-8x7B-v0.1 | 46.7B (7B active) | INT4 Q8 FP16 | 43 tok/s 27 tok/s 15 tok/s | ~629 ms ~643 ms ~673 ms | |
#220 | Ministral-8B-2410 | 8B | INT4 Q8 FP16 | 44 tok/s 26 tok/s 14 tok/s | ~617 ms ~633 ms ~667 ms | |
#221 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 146 tok/s 103 tok/s 62 tok/s | ~133 ms ~136 ms ~142 ms | |
#222 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
#223 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 80 tok/s 50 tok/s 28 tok/s | ~306 ms ~314 ms ~330 ms | |
#224 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 60 tok/s 50 tok/s 35 tok/s | ~193 ms ~196 ms ~205 ms | |
#225 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 122 tok/s 83 tok/s 48 tok/s | ~174 ms ~177 ms ~186 ms | |
#226 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 47 tok/s 28 tok/s 15 tok/s | ~565 ms ~579 ms ~611 ms | |
#227 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 181 tok/s 134 tok/s 85 tok/s | ~93 ms ~95 ms ~99 ms | |
#228 | Phi-3-medium | 14B | INT4 Q8 FP16 | 27 tok/s 15 tok/s 8 tok/s | ~1.1s ~1.1s ~1.2s | |
#229 | Yi-34B | 34B | INT4 Q8 FP16 | 9 tok/s 6 tok/s 3 tok/s | ~2.6s ~2.6s ~2.8s | |
#230 | OLMo 3 7B Instruct | 7B | INT4 Q8 FP16 | 34 tok/s 24 tok/s 14 tok/s | ~552 ms ~565 ms ~595 ms | |
#231 | Qwen2-7B | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
#232 | Phi-3-small | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
#233 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 47 tok/s 28 tok/s 15 tok/s | ~565 ms ~579 ms ~611 ms | |
#234 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 181 tok/s 134 tok/s 85 tok/s | ~93 ms ~95 ms ~99 ms | |
#235 | Falcon-180B | 180B | INT4 | 2 tok/s | ~13.1s | |
#236 | Gemma 1 7B | 7B | INT4 Q8 FP16 | 34 tok/s 24 tok/s 14 tok/s | ~552 ms ~565 ms ~595 ms | |
#237 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 132 tok/s 87 tok/s 50 tok/s | ~173 ms ~177 ms ~186 ms | |
#238 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 38 tok/s 27 tok/s 16 tok/s | ~477 ms ~488 ms ~514 ms | |
#239 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 38 tok/s 27 tok/s 16 tok/s | ~477 ms ~488 ms ~514 ms | |
#240 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 38 tok/s 27 tok/s 16 tok/s | ~477 ms ~488 ms ~514 ms | |
Unranked Models (Sorted by Speed) | ||||||
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 286 tok/s 245 tok/s 186 tok/s | ~32 ms ~32 ms ~34 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 247 tok/s 200 tok/s 140 tok/s | ~48 ms ~49 ms ~52 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 247 tok/s 200 tok/s 140 tok/s | ~48 ms ~49 ms ~52 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 241 tok/s 214 tok/s 167 tok/s | ~32 ms ~33 ms ~34 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 241 tok/s 214 tok/s 167 tok/s | ~32 ms ~33 ms ~34 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 231 tok/s 182 tok/s 124 tok/s | ~57 ms ~58 ms ~61 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 194 tok/s 141 tok/s 88 tok/s | ~93 ms ~95 ms ~99 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 181 tok/s 134 tok/s 85 tok/s | ~93 ms ~95 ms ~99 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 181 tok/s 134 tok/s 85 tok/s | ~93 ms ~95 ms ~99 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 146 tok/s 103 tok/s 62 tok/s | ~133 ms ~136 ms ~142 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 146 tok/s 103 tok/s 62 tok/s | ~133 ms ~136 ms ~142 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 136 tok/s 94 tok/s 56 tok/s | ~149 ms ~153 ms ~160 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 131 tok/s 90 tok/s 53 tok/s | ~157 ms ~160 ms ~168 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 101 tok/s 63 tok/s 35 tok/s | ~246 ms ~252 ms ~265 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 99 tok/s 81 tok/s 56 tok/s | ~121 ms ~123 ms ~128 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 99 tok/s 81 tok/s 56 tok/s | ~121 ms ~123 ms ~128 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 99 tok/s 81 tok/s 56 tok/s | ~121 ms ~123 ms ~128 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 96 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 FP16 | 96 tok/s 66 tok/s 39 tok/s | ~241 ms ~246 ms ~257 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 94 tok/s 61 tok/s 34 tok/s | ~247 ms ~253 ms ~266 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 Q8 FP16 | 87 tok/s 62 tok/s 37 tok/s | ~273 ms ~277 ms ~288 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 Q8 FP16 | 81 tok/s 55 tok/s 32 tok/s | ~301 ms ~306 ms ~319 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 Q8 FP16 | 81 tok/s 55 tok/s 32 tok/s | ~301 ms ~306 ms ~319 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 Q8 FP16 | 78 tok/s 54 tok/s 32 tok/s | ~316 ms ~321 ms ~334 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 Q8 FP16 | 78 tok/s 54 tok/s 32 tok/s | ~316 ms ~321 ms ~334 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 77 tok/s 48 tok/s 26 tok/s | ~321 ms ~329 ms ~346 ms | |
- | Hy-Memory 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 77 tok/s 54 tok/s 32 tok/s | ~321 ms ~326 ms ~339 ms | |
- | Hy-MT2 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 77 tok/s 54 tok/s 32 tok/s | ~321 ms ~326 ms ~339 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 77 tok/s 48 tok/s 26 tok/s | ~321 ms ~329 ms ~346 ms | |
- | MiniCPM-V 4.7 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 76 tok/s 53 tok/s 31 tok/s | ~327 ms ~332 ms ~345 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 73 tok/s 59 tok/s 39 tok/s | ~179 ms ~183 ms ~191 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 64 tok/s 49 tok/s 31 tok/s | ~231 ms ~236 ms ~247 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 60 tok/s 46 tok/s 29 tok/s | ~253 ms ~258 ms ~271 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 60 tok/s 46 tok/s 29 tok/s | ~253 ms ~258 ms ~271 ms | |
- | AliceAI Foundation 80B A3B Base | 80B (3B active) | INT4 Q8 | 59 tok/s 45 tok/s | ~437 ms ~443 ms | |
- | Kolibri 1 | 78B (3.5B active) | INT4 Q8 | 56 tok/s 41 tok/s | ~469 ms ~475 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 52 tok/s 30 tok/s 16 tok/s | ~542 ms ~556 ms ~586 ms | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 49 tok/s 29 tok/s 16 tok/s | ~543 ms ~557 ms ~587 ms | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 47 tok/s 28 tok/s 15 tok/s | ~565 ms ~579 ms ~611 ms | |
- | Youtu Parsing Omni | 8B | INT4 Q8 FP16 | 45 tok/s 26 tok/s 14 tok/s | ~617 ms ~632 ms ~667 ms | |
- | LensVLM 9B | 9B | INT4 Q8 FP16 | 40 tok/s 23 tok/s 12 tok/s | ~693 ms ~710 ms ~749 ms | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 FP16 | 40 tok/s 23 tok/s 12 tok/s | ~693 ms ~710 ms ~749 ms | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 38 tok/s 27 tok/s 16 tok/s | ~477 ms ~488 ms ~514 ms | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 38 tok/s 27 tok/s 16 tok/s | ~477 ms ~488 ms ~514 ms | |
- | Falcon3-10B | 10B | INT4 Q8 FP16 | 36 tok/s 21 tok/s 11 tok/s | ~767 ms ~786 ms ~829 ms | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 Q8 FP16 | 36 tok/s 31 tok/s 22 tok/s | ~304 ms ~309 ms ~322 ms | |
- | Falcon2-11B | 11B | INT4 Q8 FP16 | 35 tok/s 20 tok/s 10 tok/s | ~840 ms ~861 ms ~909 ms | |
- | Hunyuan Lite | 7B | INT4 Q8 FP16 | 34 tok/s 24 tok/s 14 tok/s | ~552 ms ~565 ms ~595 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 FP16 | 34 tok/s 24 tok/s 14 tok/s | ~552 ms ~565 ms ~595 ms | |
- | SEA-LION-7B | 7.1B | INT4 Q8 FP16 | 34 tok/s 23 tok/s 14 tok/s | ~559 ms ~573 ms ~603 ms | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 FP16 | 34 tok/s 23 tok/s 14 tok/s | ~559 ms ~573 ms ~603 ms | |
- | Laguna S 2.1 | 118B (8B active) | INT4 Q8 | 31 tok/s 20 tok/s | ~910 ms ~1.0s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 FP16 | 31 tok/s 21 tok/s 12 tok/s | ~627 ms ~642 ms ~676 ms | |
- | Typhoon-2-8B | 8B | INT4 Q8 FP16 | 31 tok/s 21 tok/s 12 tok/s | ~627 ms ~642 ms ~676 ms | |
- | GLM-4-9B | 9B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~703 ms ~720 ms ~759 ms | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~703 ms ~720 ms ~759 ms | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~703 ms ~720 ms ~759 ms | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~718 ms ~735 ms ~774 ms | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~718 ms ~735 ms ~774 ms | |
- | Yi-9B | 9B | INT4 Q8 FP16 | 28 tok/s 19 tok/s 11 tok/s | ~703 ms ~720 ms ~759 ms | |
- | GreenMind-14B-R1 | 14B | INT4 Q8 FP16 | 19 tok/s 13 tok/s 7 tok/s | ~1.1s ~1.1s ~1.2s | |
- | AREX-2 | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
- | PPLX Decider v1 27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
- | PPLX Decider v1.1 27B | 27B | INT4 Q8 FP16 | 14 tok/s 8 tok/s 4 tok/s | ~2.0s ~2.1s ~2.2s | |
- | Agnes 3.0 Flash | 33B | INT4 Q8 FP16 | 12 tok/s 7 tok/s 3 tok/s | ~2.5s ~2.5s ~2.7s | |
- | Hunyuan A13B | 80B (13B active) | INT4 Q8 | 11 tok/s 9 tok/s | ~1.2s ~1.2s | |
- | Falcon-40B | 40B | INT4 Q8 FP16 | 10 tok/s 6 tok/s 3 tok/s | ~3.0s ~3.0s ~3.2s | |
- | Typhoon-2-70B | 70B | INT4 Q8 | 5 tok/s 3 tok/s | ~5.2s ~5.4s | |
- | Kimi-Dev-72B | 72B | INT4 Q8 | 4 tok/s 3 tok/s | ~5.4s ~5.5s | |
- | GLM-130B | 130B | INT4 | 3 tok/s | ~9.7s | |
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 NVIDIA GB10 Grace Blackwell (128GB) - DGX Spark.
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 273.2 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 61 tok/s on an 8B Q4 model and 7 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning