NVIDIA A100 (80GB)
The NVIDIA A100 (80GB) features 80 GB of dedicated VRAM and 2039 GB/s memory bandwidth, supporting CUDA and TensorRT-LLM runtimes for full-precision and quantized inference.
Memory
80 GB
Usable Memory Ceiling
80 GB
Bandwidth
2039 GB/s
Max Inference Class
70B+ Models (Q4)
TDP
400W
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 A100 (80GB)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on NVIDIA A100 (80GB).
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#33 | Qwen 3.8 27B | 27B | INT4 Q8 FP16 | 91 tok/s 54 tok/s 29 tok/s | ~1.1s ~1.1s ~1.1s | |
#48 | Agnes 3.0 Flash | 33B | INT4 Q8 FP16 | 77 tok/s 45 tok/s 22 tok/s | ~1.3s ~1.3s ~1.5s | |
#61 | Sarvam-105B | 106B (10.3B active) | INT4 | 192 tok/s | ~552 ms | |
#72 | K2 Horizon MoVA 36B A4B | 36B (4B active) | INT4 Q8 FP16 | 299 tok/s 226 tok/s 138 tok/s | ~213 ms ~214 ms ~235 ms | |
#75 | Ling 3.0 Flash VL | 124B (5.5B active) | INT4 | 261 tok/s | ~428 ms | |
#83 | Qwen3.5-27B | 27B | INT4 Q8 FP16 | 91 tok/s 54 tok/s 29 tok/s | ~1.1s ~1.1s ~1.1s | |
#84 | Qwen3.5-122B-A10B | 122B (10B active) | INT4 | 147 tok/s | ~615 ms | |
#92 | Sarvam-30B | 32B (2.4B active) | INT4 Q8 FP16 | 359 tok/s 293 tok/s 194 tok/s | ~149 ms ~150 ms ~164 ms | |
#94 | Qwen3.6 35B A3B | 35B (3B active) | INT4 Q8 FP16 | 333 tok/s 263 tok/s 169 tok/s | ~174 ms ~175 ms ~192 ms | |
#96 | Muse Glimmer 30B | 30B | INT4 Q8 FP16 | 84 tok/s 49 tok/s 26 tok/s | ~1.2s ~1.2s ~1.2s | |
#100 | Phi-4 Reasoning Plus | 14B | INT4 Q8 FP16 | 117 tok/s 82 tok/s 48 tok/s | ~570 ms ~574 ms ~582 ms | |
#101 | MiMo V2 Flash | 15B (309B active) | INT4 Q8 FP16 | 9 tok/s 5 tok/s 3 tok/s | ~11.8s ~11.9s ~12.0s | |
#106 | GLM-4.5-Air | 106B (12B active) | INT4 | 57 tok/s | ~681 ms | |
#108 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 Q8 FP16 | 314 tok/s 237 tok/s 153 tok/s | ~195 ms ~196 ms ~199 ms | |
#109 | Gemma 4 12B | 11.95B | INT4 Q8 FP16 | 132 tok/s 93 tok/s 55 tok/s | ~489 ms ~492 ms ~499 ms | |
#113 | Qwen3.5-9B | 9B | INT4 Q8 FP16 | 212 tok/s 139 tok/s 79 tok/s | ~369 ms ~372 ms ~377 ms | |
#114 | Qwen3.5-35B-A3B | 35B (3B active) | INT4 Q8 FP16 | 333 tok/s 263 tok/s 169 tok/s | ~174 ms ~175 ms ~192 ms | |
#114 | Ling 3.0 Flash | 124B (5.1B active) | INT4 | 271 tok/s | ~411 ms | |
#118 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
#126 | Qwen3 Next 80B A3B | 80B (79B active) | INT4 | 28 tok/s | ~3.1s | |
#128 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 80 tok/s 48 tok/s 25 tok/s | ~1.2s ~1.2s ~1.2s | |
#129 | Gemma 4 31B | 30.7B | INT4 Q8 FP16 | 82 tok/s 48 tok/s 25 tok/s | ~1.2s ~1.2s ~1.2s | |
#130 | GPT-OSS 120B | 117B (5.1B active) | INT4 | 57 tok/s | ~452 ms | |
#131 | Nemotron 3.5 Lightning | 30B (3B active) | INT4 Q8 FP16 | 160 tok/s 147 tok/s 110 tok/s | ~174 ms ~174 ms ~191 ms | |
#132 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 337 tok/s 245 tok/s 153 tok/s | ~171 ms ~172 ms ~174 ms | |
#133 | Qwen3-30B-A3B | 30B (3B active) | INT4 Q8 FP16 | 337 tok/s 265 tok/s 178 tok/s | ~171 ms ~171 ms ~173 ms | |
#136 | GPT-OSS 20B | 21B (3.6B active) | INT4 Q8 FP16 | 175 tok/s 155 tok/s 115 tok/s | ~185 ms ~186 ms ~188 ms | |
#146 | Granite 4.2 8B | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#150 | Gemma 4 E4B | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#151 | Nemotron 3.5 Lightning 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 337 tok/s 265 tok/s 178 tok/s | ~171 ms ~171 ms ~173 ms | |
#153 | Qwen3-32B | 32B | INT4 Q8 FP16 | 79 tok/s 46 tok/s 23 tok/s | ~1.3s ~1.3s ~1.4s | |
#154 | Qwen2.5-72B | 72B | INT4 Q8 | 38 tok/s 20 tok/s | ~2.8s ~3.1s | |
#155 | Mistral Large 3 | 41B (675B active) | INT4 Q8 | 4 tok/s 2 tok/s | ~25.7s ~25.9s | |
#156 | GLM-4 | 32B | INT4 Q8 FP16 | 62 tok/s 41 tok/s 21 tok/s | ~1.3s ~1.3s ~1.4s | |
#157 | DeepSeek-R1 70B | 70B | INT4 Q8 | 40 tok/s 21 tok/s | ~2.8s ~3.0s | |
#158 | GLM-4V | 9B | INT4 Q8 FP16 | 161 tok/s 117 tok/s 71 tok/s | ~371 ms ~373 ms ~378 ms | |
#159 | Gemma 3 27B | 27B | INT4 Q8 FP16 | 91 tok/s 54 tok/s 29 tok/s | ~1.1s ~1.1s ~1.1s | |
#160 | Phi-4 | 14B | INT4 Q8 FP16 | 155 tok/s 97 tok/s 53 tok/s | ~568 ms ~572 ms ~581 ms | |
#161 | Llama 3.3 70B | 70B | INT4 Q8 | 39 tok/s 21 tok/s | ~2.8s ~3.0s | |
#162 | DeepSeek-R1 32B | 32B | INT4 Q8 FP16 | 81 tok/s 47 tok/s 23 tok/s | ~1.3s ~1.3s ~1.4s | |
#163 | Hunyuan Standard | 52B (389B active) | INT4 Q8 | 7 tok/s 4 tok/s | ~14.8s ~14.9s | |
#164 | Qwen2.5-32B | 32B | INT4 Q8 FP16 | 79 tok/s 46 tok/s 23 tok/s | ~1.3s ~1.3s ~1.4s | |
#168 | Magistral Small | 24B | INT4 Q8 FP16 | 78 tok/s 52 tok/s 29 tok/s | ~964 ms ~970 ms ~985 ms | |
#169 | OLMo 3 7B Think | 7B | INT4 Q8 FP16 | 190 tok/s 141 tok/s 88 tok/s | ~291 ms ~293 ms ~297 ms | |
#170 | OLMo 3.1 32B Instruct | 32B | INT4 Q8 FP16 | 62 tok/s 41 tok/s 21 tok/s | ~1.3s ~1.3s ~1.4s | |
#171 | Llama 3.1 70B | 70B | INT4 Q8 | 39 tok/s 21 tok/s | ~2.8s ~3.0s | |
#172 | DeepSeek-R1 14B | 14B | INT4 Q8 FP16 | 159 tok/s 98 tok/s 53 tok/s | ~568 ms ~572 ms ~580 ms | |
#173 | Ministral 3 14B | 14B | INT4 Q8 FP16 | 117 tok/s 82 tok/s 48 tok/s | ~570 ms ~574 ms ~582 ms | |
#174 | Mistral-Large-2407 | 123B | INT4 | 22 tok/s | ~5.3s | |
#175 | Llama 4 Scout | 109B (17B active) | INT4 | 116 tok/s | ~811 ms | |
#177 | Gemma 3 12B | 12B | INT4 Q8 FP16 | 174 tok/s 110 tok/s 61 tok/s | ~489 ms ~492 ms ~499 ms | |
#178 | Command A | 111B | INT4 | 19 tok/s | ~4.8s | |
#179 | Qwen2-72B | 72B | INT4 Q8 | 38 tok/s 20 tok/s | ~2.8s ~3.1s | |
#180 | OLMo 3 32B Base | 32B | INT4 Q8 FP16 | 62 tok/s 41 tok/s 21 tok/s | ~1.3s ~1.3s ~1.4s | |
#181 | Mistral-Small-2501 | 24B | INT4 Q8 FP16 | 101 tok/s 60 tok/s 32 tok/s | ~961 ms ~968 ms ~982 ms | |
#182 | Ministral 3 8B | 8B | INT4 Q8 FP16 | 175 tok/s 128 tok/s 79 tok/s | ~330 ms ~332 ms ~337 ms | |
#183 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 439 tok/s 353 tok/s 247 tok/s | ~92 ms ~93 ms ~94 ms | |
#184 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 439 tok/s 353 tok/s 247 tok/s | ~92 ms ~93 ms ~94 ms | |
#185 | Qwen2.5-14B | 14B | INT4 Q8 FP16 | 155 tok/s 97 tok/s 53 tok/s | ~568 ms ~572 ms ~581 ms | |
#186 | Gemma 2 27B | 27B | INT4 Q8 FP16 | 91 tok/s 54 tok/s 29 tok/s | ~1.1s ~1.1s ~1.1s | |
#187 | OLMo 3.1 32B Think | 32B | INT4 Q8 FP16 | 62 tok/s 41 tok/s 21 tok/s | ~1.3s ~1.3s ~1.4s | |
#188 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 298 tok/s 210 tok/s 127 tok/s | ~214 ms ~216 ms ~219 ms | |
#189 | Qwen3-4B | 4B | INT4 Q8 FP16 | 337 tok/s 245 tok/s 153 tok/s | ~171 ms ~172 ms ~174 ms | |
#190 | DeepSeek-R1 8B | 8B | INT4 Q8 FP16 | 234 tok/s 154 tok/s 88 tok/s | ~329 ms ~331 ms ~336 ms | |
#191 | Llama 3 70B | 70B | INT4 Q8 | 39 tok/s 21 tok/s | ~2.8s ~3.0s | |
#192 | Command R Plus | 104B | INT4 | 20 tok/s | ~4.5s | |
#193 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 287 tok/s 236 tok/s 164 tok/s | ~132 ms ~133 ms ~135 ms | |
#195 | Qwen3-14B | 14B | INT4 Q8 FP16 | 155 tok/s 97 tok/s 53 tok/s | ~568 ms ~572 ms ~581 ms | |
#196 | Kimi Linear 48B A3B Instruct | 48B (3B active) | INT4 Q8 | 139 tok/s 130 tok/s | ~188 ms ~189 ms | |
#197 | Gemma 2 9B | 9B | INT4 Q8 FP16 | 212 tok/s 139 tok/s 79 tok/s | ~369 ms ~372 ms ~377 ms | |
#198 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
#199 | Qwen3-8B | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#200 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 337 tok/s 245 tok/s 153 tok/s | ~171 ms ~172 ms ~174 ms | |
#201 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 254 tok/s 170 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
#202 | Llama 3.1 8B | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#204 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 462 tok/s 380 tok/s 272 tok/s | ~79 ms ~80 ms ~81 ms | |
#205 | Command R | 35B | INT4 Q8 FP16 | 58 tok/s 38 tok/s 19 tok/s | ~1.4s ~1.4s ~1.7s | |
#206 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 595 tok/s 572 tok/s 522 tok/s | ~15 ms ~15 ms ~15 ms | |
#208 | Llama 3 8B | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#209 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 578 tok/s 527 tok/s 442 tok/s | ~30 ms ~30 ms ~30 ms | |
#210 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 345 tok/s 253 tok/s 159 tok/s | ~163 ms ~164 ms ~166 ms | |
#211 | Mixtral-8x7B-v0.1 | 46.7B (7B active) | INT4 Q8 | 226 tok/s 158 tok/s | ~336 ms ~337 ms | |
#212 | Ministral-8B-2410 | 8B | INT4 Q8 FP16 | 229 tok/s 152 tok/s 87 tok/s | ~329 ms ~331 ms ~336 ms | |
#213 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 479 tok/s 400 tok/s 292 tok/s | ~70 ms ~71 ms ~72 ms | |
#214 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 381 tok/s 290 tok/s 189 tok/s | ~131 ms ~132 ms ~134 ms | |
#215 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 551 tok/s 492 tok/s 396 tok/s | ~39 ms ~39 ms ~40 ms | |
#216 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 345 tok/s 253 tok/s 159 tok/s | ~163 ms ~164 ms ~166 ms | |
#217 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 287 tok/s 254 tok/s 194 tok/s | ~99 ms ~100 ms ~101 ms | |
#218 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 243 tok/s 163 tok/s 94 tok/s | ~301 ms ~303 ms ~308 ms | |
#219 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 526 tok/s 460 tok/s 359 tok/s | ~49 ms ~49 ms ~50 ms | |
#220 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 439 tok/s 353 tok/s 247 tok/s | ~92 ms ~93 ms ~94 ms | |
#221 | Phi-3-medium | 14B | INT4 Q8 FP16 | 155 tok/s 97 tok/s 53 tok/s | ~568 ms ~572 ms ~581 ms | |
#222 | OLMo 3 7B Instruct | 7B | INT4 Q8 FP16 | 190 tok/s 141 tok/s 88 tok/s | ~291 ms ~293 ms ~297 ms | |
#223 | Yi-34B | 34B | INT4 Q8 FP16 | 60 tok/s 38 tok/s 20 tok/s | ~1.4s ~1.4s ~1.5s | |
#224 | Qwen2-7B | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
#225 | Phi-3-small | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
#226 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 526 tok/s 460 tok/s 359 tok/s | ~49 ms ~49 ms ~50 ms | |
#227 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 243 tok/s 163 tok/s 94 tok/s | ~301 ms ~303 ms ~308 ms | |
#229 | Gemma 1 7B | 7B | INT4 Q8 FP16 | 190 tok/s 141 tok/s 88 tok/s | ~291 ms ~293 ms ~297 ms | |
#230 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 456 tok/s 364 tok/s 251 tok/s | ~92 ms ~93 ms ~94 ms | |
#231 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 208 tok/s 156 tok/s 99 tok/s | ~251 ms ~253 ms ~256 ms | |
#232 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 208 tok/s 156 tok/s 99 tok/s | ~251 ms ~253 ms ~256 ms | |
#233 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 208 tok/s 156 tok/s 99 tok/s | ~251 ms ~253 ms ~256 ms | |
Unranked Models (Sorted by Speed) | ||||||
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 591 tok/s 547 tok/s 468 tok/s | ~25 ms ~25 ms ~26 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 591 tok/s 547 tok/s 468 tok/s | ~25 ms ~25 ms ~26 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 586 tok/s 561 tok/s 509 tok/s | ~16 ms ~17 ms ~17 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 586 tok/s 561 tok/s 509 tok/s | ~16 ms ~17 ms ~17 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 541 tok/s 470 tok/s 365 tok/s | ~49 ms ~49 ms ~50 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 526 tok/s 460 tok/s 359 tok/s | ~49 ms ~49 ms ~50 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 479 tok/s 400 tok/s 292 tok/s | ~70 ms ~71 ms ~72 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 479 tok/s 400 tok/s 292 tok/s | ~70 ms ~71 ms ~72 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 462 tok/s 380 tok/s 272 tok/s | ~79 ms ~80 ms ~81 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 454 tok/s 371 tok/s 263 tok/s | ~83 ms ~84 ms ~85 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 395 tok/s 297 tok/s 192 tok/s | ~131 ms ~132 ms ~134 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 391 tok/s 348 tok/s 273 tok/s | ~62 ms ~63 ms ~63 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 391 tok/s 348 tok/s 273 tok/s | ~62 ms ~63 ms ~63 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 391 tok/s 348 tok/s 273 tok/s | ~62 ms ~63 ms ~63 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 385 tok/s 291 tok/s 190 tok/s | ~131 ms ~132 ms ~134 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 381 tok/s 290 tok/s 189 tok/s | ~131 ms ~132 ms ~134 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 381 tok/s 290 tok/s 189 tok/s | ~131 ms ~132 ms ~134 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 Q8 FP16 | 348 tok/s 271 tok/s 181 tok/s | ~160 ms ~161 ms ~163 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 Q8 FP16 | 348 tok/s 271 tok/s 181 tok/s | ~160 ms ~161 ms ~163 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 Q8 FP16 | 339 tok/s 267 tok/s 179 tok/s | ~168 ms ~169 ms ~171 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 Q8 FP16 | 339 tok/s 267 tok/s 179 tok/s | ~168 ms ~169 ms ~171 ms | |
- | Hy-MT2 30B A3B | 30B (3B active) | INT4 Q8 FP16 | 337 tok/s 265 tok/s 178 tok/s | ~171 ms ~171 ms ~173 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 337 tok/s 245 tok/s 153 tok/s | ~171 ms ~172 ms ~174 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 298 tok/s 248 tok/s 176 tok/s | ~120 ms ~121 ms ~123 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 287 tok/s 236 tok/s 164 tok/s | ~132 ms ~133 ms ~135 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 287 tok/s 236 tok/s 164 tok/s | ~132 ms ~133 ms ~135 ms | |
- | AliceAI Foundation 80B A3B Base | 80B (3B active) | INT4 Q8 | 284 tok/s 210 tok/s | ~233 ms ~275 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 258 tok/s 172 tok/s 99 tok/s | ~289 ms ~291 ms ~296 ms | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 249 tok/s 168 tok/s 98 tok/s | ~289 ms ~291 ms ~296 ms | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 243 tok/s 163 tok/s 94 tok/s | ~301 ms ~303 ms ~308 ms | |
- | LensVLM 9B | 9B | INT4 Q8 FP16 | 212 tok/s 139 tok/s 79 tok/s | ~369 ms ~372 ms ~377 ms | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 FP16 | 212 tok/s 139 tok/s 79 tok/s | ~369 ms ~372 ms ~377 ms | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 208 tok/s 156 tok/s 99 tok/s | ~251 ms ~253 ms ~256 ms | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 208 tok/s 156 tok/s 99 tok/s | ~251 ms ~253 ms ~256 ms | |
- | Falcon3-10B | 10B | INT4 Q8 FP16 | 198 tok/s 128 tok/s 72 tok/s | ~409 ms ~411 ms ~418 ms | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 Q8 FP16 | 198 tok/s 176 tok/s 133 tok/s | ~156 ms ~156 ms ~158 ms | |
- | Falcon2-11B | 11B | INT4 Q8 FP16 | 191 tok/s 121 tok/s 66 tok/s | ~448 ms ~451 ms ~458 ms | |
- | Hunyuan Lite | 7B | INT4 Q8 FP16 | 190 tok/s 141 tok/s 88 tok/s | ~291 ms ~293 ms ~297 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 FP16 | 190 tok/s 141 tok/s 88 tok/s | ~291 ms ~293 ms ~297 ms | |
- | SEA-LION-7B | 7.1B | INT4 Q8 FP16 | 189 tok/s 139 tok/s 87 tok/s | ~295 ms ~296 ms ~301 ms | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 FP16 | 189 tok/s 139 tok/s 87 tok/s | ~295 ms ~296 ms ~301 ms | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 FP16 | 175 tok/s 128 tok/s 79 tok/s | ~330 ms ~332 ms ~337 ms | |
- | Typhoon-2-8B | 8B | INT4 Q8 FP16 | 175 tok/s 128 tok/s 79 tok/s | ~330 ms ~332 ms ~337 ms | |
- | Laguna S 2.1 | 118B (8B active) | INT4 | 174 tok/s | ~486 ms | |
- | GLM-4-9B | 9B | INT4 Q8 FP16 | 161 tok/s 117 tok/s 71 tok/s | ~371 ms ~373 ms ~378 ms | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 FP16 | 161 tok/s 117 tok/s 71 tok/s | ~371 ms ~373 ms ~378 ms | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 FP16 | 161 tok/s 117 tok/s 71 tok/s | ~371 ms ~373 ms ~378 ms | |
- | Yi-9B | 9B | INT4 Q8 FP16 | 161 tok/s 117 tok/s 71 tok/s | ~371 ms ~373 ms ~378 ms | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 FP16 | 159 tok/s 115 tok/s 70 tok/s | ~378 ms ~381 ms ~386 ms | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 FP16 | 159 tok/s 115 tok/s 70 tok/s | ~378 ms ~381 ms ~386 ms | |
- | GreenMind-14B-R1 | 14B | INT4 Q8 FP16 | 117 tok/s 82 tok/s 48 tok/s | ~570 ms ~574 ms ~582 ms | |
- | Hunyuan A13B | 80B (13B active) | INT4 | 71 tok/s | ~620 ms | |
- | Falcon-40B | 40B | INT4 Q8 FP16 | 67 tok/s 38 tok/s 18 tok/s | ~1.6s ~1.6s ~1.9s | |
- | Typhoon-2-70B | 70B | INT4 Q8 | 31 tok/s 18 tok/s | ~2.8s ~3.3s | |
- | Kimi-Dev-72B | 72B | INT4 Q8 | 30 tok/s 17 tok/s | ~2.8s ~3.4s | |
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 A100 (80GB).
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 2039 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 453 tok/s on an 8B Q4 model and 54 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning