NVIDIA H100 NVL (94GB)
The NVIDIA H100 NVL (94GB) features 94 GB of dedicated VRAM and 3938 GB/s memory bandwidth, supporting CUDA and TensorRT-LLM runtimes for full-precision and quantized inference.
Memory
94 GB
Usable Memory Ceiling
94 GB
Bandwidth
3938 GB/s
Max Inference Class
70B+ Models (Q4)
TDP
700W
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 H100 NVL (94GB)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on NVIDIA H100 NVL (94GB).
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#33 | Qwen 3.8 27B | 27B | INT4 Q8 FP16 | 157 tok/s 97 tok/s 53 tok/s | ~408 ms ~412 ms ~420 ms | |
#41 | Qwen 3.8 Flash | 176B (6B active) | INT4 | 351 tok/s | ~106 ms | |
#48 | Agnes 3.0 Flash | 33B | INT4 Q8 FP16 | 134 tok/s 82 tok/s 44 tok/s | ~495 ms ~500 ms ~511 ms | |
#61 | Sarvam-105B | 106B (10.3B active) | INT4 | 292 tok/s | ~210 ms | |
#72 | K2 Horizon MoVA 36B A4B | 36B (4B active) | INT4 Q8 FP16 | 407 tok/s 332 tok/s 233 tok/s | ~82 ms ~83 ms ~84 ms | |
#75 | Ling 3.0 Flash VL | 124B (5.5B active) | INT4 | 384 tok/s | ~150 ms | |
#83 | Qwen3.5-27B | 27B | INT4 Q8 FP16 | 157 tok/s 97 tok/s 53 tok/s | ~408 ms ~412 ms ~420 ms | |
#84 | Qwen3.5-122B-A10B | 122B (10B active) | INT4 | 246 tok/s | ~215 ms | |
#92 | Sarvam-30B | 32B (2.4B active) | INT4 Q8 FP16 | 462 tok/s 401 tok/s 307 tok/s | ~58 ms ~59 ms ~59 ms | |
#94 | Qwen3.6 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 439 tok/s 372 tok/s 275 tok/s | ~68 ms ~68 ms ~69 ms | |
#96 | Muse Glimmer 30B | 30B | INT4 Q8 FP16 | 145 tok/s 89 tok/s 48 tok/s | ~452 ms ~456 ms ~466 ms | |
#100 | Phi-4 Reasoning Plus | 14B | INT4 Q8 FP16 | 195 tok/s 142 tok/s 87 tok/s | ~217 ms ~219 ms ~223 ms | |
#101 | MiMo V2 Flash | 15B (309B active) | INT4 Q8 FP16 | 18 tok/s 10 tok/s 5 tok/s | ~4.4s ~4.5s ~4.6s | |
#106 | GLM-4.5-Air | 106B (12B active) | INT4 | 107 tok/s | ~239 ms | |
#108 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 Q8 FP16 | 421 tok/s 344 tok/s 243 tok/s | ~76 ms ~76 ms ~77 ms | |
#109 | Gemma 4 12B | 11.95B | INT4 Q8 FP16 | 215 tok/s 159 tok/s 99 tok/s | ~186 ms ~188 ms ~192 ms | |
#113 | Qwen3.5-9B | 9B | INT4 Q8 FP16 | 316 tok/s 224 tok/s 137 tok/s | ~141 ms ~142 ms ~145 ms | |
#114 | Qwen3.5-35B-A3B | 35B (3B active) | INT4 Q8 FP16 | 439 tok/s 372 tok/s 275 tok/s | ~68 ms ~68 ms ~69 ms | |
#114 | Ling 3.0 Flash | 124B (5.1B active) | INT4 | 394 tok/s | ~144 ms | |
#118 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
#126 | Qwen3 Next 80B A3B | 80B (79B active) | INT4 Q8 | 52 tok/s 30 tok/s | ~1.2s ~1.4s | |
#128 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 139 tok/s 87 tok/s 48 tok/s | ~438 ms ~442 ms ~452 ms | |
#129 | Gemma 4 31B | 30.7B | INT4 Q8 FP16 | 142 tok/s 87 tok/s 47 tok/s | ~462 ms ~466 ms ~476 ms | |
#130 | GPT-OSS 120B | 117B (5.1B active) | INT4 | 107 tok/s | ~161 ms | |
#131 | Nemotron 3.5 Lightning | 30B (3B active) | INT4 Q8 FP16 | 252 tok/s 236 tok/s 193 tok/s | ~68 ms ~68 ms ~69 ms | |
#132 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 442 tok/s 353 tok/s 244 tok/s | ~66 ms ~67 ms ~68 ms | |
#133 | Qwen3-30B-A3B | 30B (3B active) | INT4 Q8 FP16 | 442 tok/s 374 tok/s 276 tok/s | ~66 ms ~67 ms ~68 ms | |
#136 | GPT-OSS 20B | 21B (3.6B active) | INT4 Q8 FP16 | 272 tok/s 246 tok/s 192 tok/s | ~72 ms ~73 ms ~74 ms | |
#144 | Mistral Medium 3.5 128B | 128B | INT4 | 31 tok/s | ~2.1s | |
#146 | Granite 4.2 8B | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#150 | Gemma 4 E4B | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#151 | Nemotron 3.5 Lightning 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 442 tok/s 374 tok/s 276 tok/s | ~66 ms ~67 ms ~68 ms | |
#153 | Qwen3-32B | 32B | INT4 Q8 FP16 | 138 tok/s 84 tok/s 45 tok/s | ~481 ms ~486 ms ~496 ms | |
#154 | Qwen2.5-72B | 72B | INT4 Q8 | 69 tok/s 40 tok/s | ~1.1s ~1.1s | |
#155 | Mistral Large 3 | 41B (675B active) | INT4 Q8 FP16 | 8 tok/s 5 tok/s 2 tok/s | ~9.7s ~9.8s ~11.7s | |
#156 | GLM-4 | 32B | INT4 Q8 FP16 | 111 tok/s 74 tok/s 42 tok/s | ~483 ms ~487 ms ~497 ms | |
#157 | DeepSeek-R1 70B | 70B | INT4 Q8 | 73 tok/s 42 tok/s | ~1.0s ~1.1s | |
#158 | GLM-4V | 9B | INT4 Q8 FP16 | 254 tok/s 194 tok/s 125 tok/s | ~142 ms ~143 ms ~146 ms | |
#159 | Gemma 3 27B | 27B | INT4 Q8 FP16 | 157 tok/s 97 tok/s 53 tok/s | ~408 ms ~412 ms ~420 ms | |
#160 | Phi-4 | 14B | INT4 Q8 FP16 | 246 tok/s 164 tok/s 95 tok/s | ~216 ms ~218 ms ~222 ms | |
#161 | Llama 3.3 70B | 70B | INT4 Q8 | 71 tok/s 41 tok/s | ~1.0s ~1.1s | |
#162 | DeepSeek-R1 32B | 32B | INT4 Q8 FP16 | 141 tok/s 85 tok/s 46 tok/s | ~481 ms ~485 ms ~496 ms | |
#163 | Hunyuan Standard | 52B (389B active) | INT4 Q8 | 14 tok/s 8 tok/s | ~5.6s ~5.6s | |
#164 | Qwen2.5-32B | 32B | INT4 Q8 FP16 | 138 tok/s 84 tok/s 45 tok/s | ~481 ms ~486 ms ~496 ms | |
#168 | Magistral Small | 24B | INT4 Q8 FP16 | 136 tok/s 94 tok/s 55 tok/s | ~365 ms ~369 ms ~376 ms | |
#169 | OLMo 3 7B Think | 7B | INT4 Q8 FP16 | 290 tok/s 227 tok/s 151 tok/s | ~112 ms ~113 ms ~115 ms | |
#170 | OLMo 3.1 32B Instruct | 32B | INT4 Q8 FP16 | 111 tok/s 74 tok/s 42 tok/s | ~483 ms ~487 ms ~497 ms | |
#171 | Llama 3.1 70B | 70B | INT4 Q8 | 71 tok/s 41 tok/s | ~1.0s ~1.1s | |
#172 | DeepSeek-R1 14B | 14B | INT4 Q8 FP16 | 251 tok/s 167 tok/s 96 tok/s | ~216 ms ~218 ms ~222 ms | |
#173 | Ministral 3 14B | 14B | INT4 Q8 FP16 | 195 tok/s 142 tok/s 87 tok/s | ~217 ms ~219 ms ~223 ms | |
#174 | Mistral-Large-2407 | 123B | INT4 | 42 tok/s | ~1.8s | |
#175 | Llama 4 Scout | 109B (17B active) | INT4 | 193 tok/s | ~307 ms | |
#177 | Gemma 3 12B | 12B | INT4 Q8 FP16 | 270 tok/s 184 tok/s 108 tok/s | ~186 ms ~188 ms ~192 ms | |
#178 | Command A | 111B | INT4 | 36 tok/s | ~1.8s | |
#179 | Qwen2-72B | 72B | INT4 Q8 | 69 tok/s 40 tok/s | ~1.1s ~1.1s | |
#180 | OLMo 3 32B Base | 32B | INT4 Q8 FP16 | 111 tok/s 74 tok/s 42 tok/s | ~483 ms ~487 ms ~497 ms | |
#181 | Mistral-Small-2501 | 24B | INT4 Q8 FP16 | 171 tok/s 107 tok/s 59 tok/s | ~364 ms ~367 ms ~375 ms | |
#182 | Ministral 3 8B | 8B | INT4 Q8 FP16 | 272 tok/s 210 tok/s 137 tok/s | ~127 ms ~128 ms ~130 ms | |
#183 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 526 tok/s 457 tok/s 354 tok/s | ~37 ms ~37 ms ~38 ms | |
#184 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 526 tok/s 457 tok/s 354 tok/s | ~37 ms ~37 ms ~38 ms | |
#185 | Qwen2.5-14B | 14B | INT4 Q8 FP16 | 246 tok/s 164 tok/s 95 tok/s | ~216 ms ~218 ms ~222 ms | |
#186 | Gemma 2 27B | 27B | INT4 Q8 FP16 | 157 tok/s 97 tok/s 53 tok/s | ~408 ms ~412 ms ~420 ms | |
#187 | OLMo 3.1 32B Think | 32B | INT4 Q8 FP16 | 111 tok/s 74 tok/s 42 tok/s | ~483 ms ~487 ms ~497 ms | |
#188 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 407 tok/s 313 tok/s 208 tok/s | ~83 ms ~84 ms ~85 ms | |
#189 | Qwen3-4B | 4B | INT4 Q8 FP16 | 442 tok/s 353 tok/s 244 tok/s | ~66 ms ~67 ms ~68 ms | |
#190 | DeepSeek-R1 8B | 8B | INT4 Q8 FP16 | 341 tok/s 245 tok/s 151 tok/s | ~126 ms ~127 ms ~130 ms | |
#191 | Llama 3 70B | 70B | INT4 Q8 | 71 tok/s 41 tok/s | ~1.0s ~1.1s | |
#192 | Command R Plus | 104B | INT4 | 40 tok/s | ~1.6s | |
#193 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 396 tok/s 342 tok/s 258 tok/s | ~52 ms ~52 ms ~53 ms | |
#195 | Qwen3-14B | 14B | INT4 Q8 FP16 | 246 tok/s 164 tok/s 95 tok/s | ~216 ms ~218 ms ~222 ms | |
#196 | Kimi Linear 48B A3B Instruct | 48B (3B active) | INT4 Q8 | 225 tok/s 213 tok/s | ~74 ms ~74 ms | |
#197 | Gemma 2 9B | 9B | INT4 Q8 FP16 | 316 tok/s 224 tok/s 137 tok/s | ~141 ms ~142 ms ~145 ms | |
#198 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
#199 | Qwen3-8B | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#200 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 442 tok/s 353 tok/s 244 tok/s | ~66 ms ~67 ms ~68 ms | |
#201 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 362 tok/s 265 tok/s 167 tok/s | ~111 ms ~112 ms ~114 ms | |
#202 | Llama 3.1 8B | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#203 | Mixtral-8x22B-v0.1 | 176B (22B active) | INT4 | 147 tok/s | ~453 ms | |
#204 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 542 tok/s 479 tok/s 381 tok/s | ~32 ms ~32 ms ~33 ms | |
#205 | Command R | 35B | INT4 Q8 FP16 | 104 tok/s 69 tok/s 37 tok/s | ~526 ms ~531 ms ~588 ms | |
#206 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 627 tok/s 614 tok/s 583 tok/s | ~8 ms ~8 ms ~8 ms | |
#208 | Llama 3 8B | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#209 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 617 tok/s 586 tok/s 527 tok/s | ~13 ms ~13 ms ~14 ms | |
#210 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 449 tok/s 361 tok/s 251 tok/s | ~63 ms ~64 ms ~65 ms | |
#211 | Mixtral-8x7B-v0.1 | 46.7B (7B active) | INT4 Q8 FP16 | 332 tok/s 250 tok/s 145 tok/s | ~128 ms ~129 ms ~154 ms | |
#212 | Ministral-8B-2410 | 8B | INT4 Q8 FP16 | 336 tok/s 242 tok/s 150 tok/s | ~126 ms ~127 ms ~130 ms | |
#213 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 554 tok/s 496 tok/s 401 tok/s | ~29 ms ~29 ms ~29 ms | |
#214 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 480 tok/s 398 tok/s 289 tok/s | ~52 ms ~52 ms ~53 ms | |
#215 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 601 tok/s 563 tok/s 492 tok/s | ~17 ms ~17 ms ~17 ms | |
#216 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 449 tok/s 361 tok/s 251 tok/s | ~63 ms ~64 ms ~65 ms | |
#217 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 396 tok/s 362 tok/s 294 tok/s | ~40 ms ~40 ms ~41 ms | |
#218 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 350 tok/s 256 tok/s 161 tok/s | ~115 ms ~116 ms ~119 ms | |
#219 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 586 tok/s 541 tok/s 462 tok/s | ~21 ms ~21 ms ~21 ms | |
#220 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 526 tok/s 457 tok/s 354 tok/s | ~37 ms ~37 ms ~38 ms | |
#221 | Phi-3-medium | 14B | INT4 Q8 FP16 | 246 tok/s 164 tok/s 95 tok/s | ~216 ms ~218 ms ~222 ms | |
#222 | OLMo 3 7B Instruct | 7B | INT4 Q8 FP16 | 290 tok/s 227 tok/s 151 tok/s | ~112 ms ~113 ms ~115 ms | |
#223 | Yi-34B | 34B | INT4 Q8 FP16 | 106 tok/s 70 tok/s 40 tok/s | ~512 ms ~517 ms ~528 ms | |
#224 | Qwen2-7B | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
#225 | Phi-3-small | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
#226 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 586 tok/s 541 tok/s 462 tok/s | ~21 ms ~21 ms ~21 ms | |
#227 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 350 tok/s 256 tok/s 161 tok/s | ~115 ms ~116 ms ~119 ms | |
#228 | Falcon-180B | 180B | INT4 | 27 tok/s | ~3.1s | |
#229 | Gemma 1 7B | 7B | INT4 Q8 FP16 | 290 tok/s 227 tok/s 151 tok/s | ~112 ms ~113 ms ~115 ms | |
#230 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 538 tok/s 466 tok/s 359 tok/s | ~37 ms ~37 ms ~38 ms | |
#231 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 311 tok/s 248 tok/s 168 tok/s | ~97 ms ~98 ms ~99 ms | |
#232 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 311 tok/s 248 tok/s 168 tok/s | ~97 ms ~98 ms ~99 ms | |
#233 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 311 tok/s 248 tok/s 168 tok/s | ~97 ms ~98 ms ~99 ms | |
Unranked Models (Sorted by Speed) | ||||||
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 625 tok/s 599 tok/s 547 tok/s | ~12 ms ~12 ms ~12 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 625 tok/s 599 tok/s 547 tok/s | ~12 ms ~12 ms ~12 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 622 tok/s 608 tok/s 574 tok/s | ~8 ms ~8 ms ~8 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 622 tok/s 608 tok/s 574 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 595 tok/s 548 tok/s 467 tok/s | ~21 ms ~21 ms ~21 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 586 tok/s 541 tok/s 462 tok/s | ~21 ms ~21 ms ~21 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 554 tok/s 496 tok/s 401 tok/s | ~29 ms ~29 ms ~29 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 554 tok/s 496 tok/s 401 tok/s | ~29 ms ~29 ms ~29 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 542 tok/s 479 tok/s 381 tok/s | ~32 ms ~32 ms ~33 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 537 tok/s 472 tok/s 372 tok/s | ~34 ms ~34 ms ~34 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 492 tok/s 405 tok/s 293 tok/s | ~51 ms ~52 ms ~53 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 488 tok/s 452 tok/s 381 tok/s | ~26 ms ~26 ms ~26 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 488 tok/s 452 tok/s 381 tok/s | ~26 ms ~26 ms ~26 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 488 tok/s 452 tok/s 381 tok/s | ~26 ms ~26 ms ~26 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 483 tok/s 400 tok/s 290 tok/s | ~52 ms ~52 ms ~53 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 480 tok/s 398 tok/s 289 tok/s | ~52 ms ~52 ms ~53 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 480 tok/s 398 tok/s 289 tok/s | ~52 ms ~52 ms ~53 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 Q8 FP16 | 452 tok/s 380 tok/s 279 tok/s | ~62 ms ~63 ms ~64 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 Q8 FP16 | 452 tok/s 380 tok/s 279 tok/s | ~62 ms ~63 ms ~64 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 Q8 FP16 | 445 tok/s 375 tok/s 276 tok/s | ~65 ms ~66 ms ~67 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 Q8 FP16 | 445 tok/s 375 tok/s 276 tok/s | ~65 ms ~66 ms ~67 ms | |
- | Hy-MT2 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 442 tok/s 374 tok/s 276 tok/s | ~66 ms ~67 ms ~68 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 442 tok/s 353 tok/s 244 tok/s | ~66 ms ~67 ms ~68 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 406 tok/s 356 tok/s 272 tok/s | ~48 ms ~48 ms ~49 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 396 tok/s 342 tok/s 258 tok/s | ~52 ms ~52 ms ~53 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 396 tok/s 342 tok/s 258 tok/s | ~52 ms ~52 ms ~53 ms | |
- | AliceAI Foundation 80B A3B Base | 80B (3B active) | INT4 Q8 | 393 tok/s 323 tok/s | ~90 ms ~98 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 366 tok/s 267 tok/s 168 tok/s | ~111 ms ~112 ms ~114 ms | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 357 tok/s 263 tok/s 166 tok/s | ~111 ms ~112 ms ~114 ms | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 350 tok/s 256 tok/s 161 tok/s | ~115 ms ~116 ms ~119 ms | |
- | LensVLM 9B | 9B | INT4 Q8 FP16 | 316 tok/s 224 tok/s 137 tok/s | ~141 ms ~142 ms ~145 ms | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 FP16 | 316 tok/s 224 tok/s 137 tok/s | ~141 ms ~142 ms ~145 ms | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 311 tok/s 248 tok/s 168 tok/s | ~97 ms ~98 ms ~99 ms | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 311 tok/s 248 tok/s 168 tok/s | ~97 ms ~98 ms ~99 ms | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 Q8 FP16 | 300 tok/s 273 tok/s 217 tok/s | ~61 ms ~61 ms ~62 ms | |
- | Falcon3-10B | 10B | INT4 Q8 FP16 | 299 tok/s 209 tok/s 126 tok/s | ~156 ms ~157 ms ~161 ms | |
- | Falcon2-11B | 11B | INT4 Q8 FP16 | 292 tok/s 199 tok/s 117 tok/s | ~171 ms ~172 ms ~176 ms | |
- | Hunyuan Lite | 7B | INT4 Q8 FP16 | 290 tok/s 227 tok/s 151 tok/s | ~112 ms ~113 ms ~115 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 FP16 | 290 tok/s 227 tok/s 151 tok/s | ~112 ms ~113 ms ~115 ms | |
- | SEA-LION-7B | 7.1B | INT4 Q8 FP16 | 288 tok/s 225 tok/s 149 tok/s | ~113 ms ~114 ms ~116 ms | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 FP16 | 288 tok/s 225 tok/s 149 tok/s | ~113 ms ~114 ms ~116 ms | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 FP16 | 272 tok/s 210 tok/s 137 tok/s | ~127 ms ~128 ms ~130 ms | |
- | Typhoon-2-8B | 8B | INT4 Q8 FP16 | 272 tok/s 210 tok/s 137 tok/s | ~127 ms ~128 ms ~130 ms | |
- | Laguna S 2.1 | 118B (8B active) | INT4 | 271 tok/s | ~185 ms | |
- | GLM-4-9B | 9B | INT4 Q8 FP16 | 254 tok/s 194 tok/s 125 tok/s | ~142 ms ~143 ms ~146 ms | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 FP16 | 254 tok/s 194 tok/s 125 tok/s | ~142 ms ~143 ms ~146 ms | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 FP16 | 254 tok/s 194 tok/s 125 tok/s | ~142 ms ~143 ms ~146 ms | |
- | Yi-9B | 9B | INT4 Q8 FP16 | 254 tok/s 194 tok/s 125 tok/s | ~142 ms ~143 ms ~146 ms | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 FP16 | 252 tok/s 191 tok/s 123 tok/s | ~145 ms ~146 ms ~149 ms | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 FP16 | 252 tok/s 191 tok/s 123 tok/s | ~145 ms ~146 ms ~149 ms | |
- | GreenMind-14B-R1 | 14B | INT4 Q8 FP16 | 195 tok/s 142 tok/s 87 tok/s | ~217 ms ~219 ms ~223 ms | |
- | Hunyuan A13B | 80B (13B active) | INT4 Q8 | 125 tok/s 95 tok/s | ~236 ms ~279 ms | |
- | Falcon-40B | 40B | INT4 Q8 FP16 | 118 tok/s 70 tok/s 35 tok/s | ~597 ms ~603 ms ~668 ms | |
- | Typhoon-2-70B | 70B | INT4 Q8 | 58 tok/s 35 tok/s | ~1.0s ~1.1s | |
- | Kimi-Dev-72B | 72B | INT4 Q8 | 56 tok/s 34 tok/s | ~1.1s ~1.2s | |
- | GLM-130B | 130B | INT4 | 31 tok/s | ~2.1s | |
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 H100 NVL (94GB).
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 3938 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 875 tok/s on an 8B Q4 model and 104 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning