NVIDIA RTX 6000 Ada Generation (48GB)
The NVIDIA RTX 6000 Ada Generation (48GB) features 48 GB of dedicated VRAM and 960 GB/s memory bandwidth, supporting CUDA and TensorRT-LLM runtimes for full-precision and quantized inference.
Memory
48 GB
Usable Memory Ceiling
48 GB
Bandwidth
960 GB/s
Max Inference Class
32B Models (Q4)
TDP
300W
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 RTX 6000 Ada Generation (48GB)
Inference throughput, memory compatibility, and generation speeds for open-weights LLMs on NVIDIA RTX 6000 Ada Generation (48GB).
Sort by:
Context:
KV Cache:
GPUs:
| Rank | Model | Parameters | Precision | TPS (Approx.) | TTFT | Can Run |
|---|---|---|---|---|---|---|
#36 | Qwen 3.8 27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
#80 | K2 Horizon MoVA 36B A4B | 36B (4B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~10 ms ~10 ms | |
#86 | Qwen3.5-27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
#93 | Sarvam-30B | 32B (2.4B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#99 | Muse Glimmer 30B | 30B | INT4 Q8 | 370 tok/s 370 tok/s | ~23 ms ~23 ms | |
#102 | Phi-4 Reasoning Plus | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#103 | Qwen3.6 35B A3B | 35B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#105 | MiMo V2 Flash | 15B (309B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~162 ms ~162 ms ~162 ms | |
#110 | Gemma 4 26B A4B | 25.2B (3.8B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~10 ms ~10 ms | |
#117 | Nemotron 3.5 Bolt 74B A9B | 74B | INT4 | 352 tok/s | ~45 ms | |
#120 | Qwen3.5-35B-A3B | 35B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#121 | Qwen3.5-9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#127 | K2 Horizon 7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#130 | Gemma 4 12B | 11.95B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~14 ms ~14 ms ~14 ms | |
#136 | Qwen3.5-4B | 4B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#139 | GPT-OSS 20B | 21B (3.6B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 333 tok/s | ~10 ms ~10 ms ~10 ms | |
#140 | GLM-4V | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#145 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~23 ms ~23 ms ~23 ms | |
#150 | Qwen3-30B-A3B | 30B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#152 | Granite 4.2 8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#157 | Gemma 4 E4B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#158 | Mistral Large 3 | 41B (675B active) | INT4 | 370 tok/s | ~345 ms | |
#159 | Qwen2.5-72B | 72B | INT4 | 352 tok/s | ~44 ms | |
#161 | GLM-4 | 32B | INT4 Q8 | 370 tok/s 352 tok/s | ~24 ms ~24 ms | |
#162 | DeepSeek-R1 70B | 70B | INT4 | 370 tok/s | ~43 ms | |
#163 | Gemma 3 27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
#164 | Phi-4 | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#165 | Llama 3.3 70B | 70B | INT4 | 370 tok/s | ~43 ms | |
#166 | Qwen3-32B | 32B | INT4 Q8 | 370 tok/s 370 tok/s | ~24 ms ~24 ms | |
#167 | Hunyuan Standard | 52B (389B active) | INT4 | 370 tok/s | ~202 ms | |
#168 | DeepSeek-R1 32B | 32B | INT4 Q8 | 370 tok/s 370 tok/s | ~24 ms ~24 ms | |
#169 | Gemma 4 31B | 30.7B | INT4 Q8 | 370 tok/s 370 tok/s | ~23 ms ~23 ms | |
#171 | Magistral Small | 24B | INT4 Q8 | 370 tok/s 370 tok/s | ~20 ms ~20 ms | |
#172 | Qwen2.5-32B | 32B | INT4 Q8 | 370 tok/s 370 tok/s | ~24 ms ~24 ms | |
#174 | OLMo 3.1 32B Instruct | 32B | INT4 Q8 | 370 tok/s 352 tok/s | ~24 ms ~24 ms | |
#175 | Llama 3.1 70B | 70B | INT4 | 370 tok/s | ~43 ms | |
#176 | DeepSeek-R1 14B | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#177 | Ministral 3 14B | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#180 | Gemma 3 12B | 12B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~14 ms ~14 ms ~14 ms | |
#183 | Qwen2-72B | 72B | INT4 | 352 tok/s | ~44 ms | |
#184 | Mistral-Small-2501 | 24B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 333 tok/s | ~20 ms ~20 ms ~20 ms | |
#185 | Ministral 3 8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#186 | Gemma 2 27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
#187 | OLMo 3 7B Think | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#188 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#189 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#190 | Qwen2.5-14B | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#191 | OLMo 3.1 32B Think | 32B | INT4 Q8 | 370 tok/s 352 tok/s | ~24 ms ~24 ms | |
#192 | Llama 3 70B | 70B | INT4 | 370 tok/s | ~43 ms | |
#193 | Gemma 4 E2B | 5.1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#195 | Ministral 3 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#196 | DeepSeek-R1 8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#198 | Gemma 2 9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#199 | Nemotron 3.5 Lightning 30B A3B | 30B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#200 | Kimi Linear 48B A3B Instruct | 48B (3B active) | INT4 | 370 tok/s | ~9 ms | |
#201 | Qwen3-4B | 4B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#202 | Qwen3-14B | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#203 | Gemma 3 4B | 4B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#204 | DeepSeek-R1 7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#205 | OLMo 3 32B Base | 32B | INT4 Q8 | 370 tok/s 352 tok/s | ~24 ms ~24 ms | |
#206 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#207 | Llama 3.1 8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#207 | Qwen3-8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#210 | Nemotron 3.5 Lightning | 30B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
#211 | Command R | 35B | INT4 Q8 | 370 tok/s 352 tok/s | ~25 ms ~25 ms | |
#212 | Llama 3 8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#213 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#214 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#216 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#217 | Phi-4-Mini | 3.8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#218 | Qwen2.5-7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#219 | Mixtral-8x7B-v0.1 | 46.7B (7B active) | INT4 Q8 | 370 tok/s 333 tok/s | ~11 ms ~12 ms | |
#220 | Ministral-8B-2410 | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
#221 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#222 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#223 | Phi-3-mini | 3.8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
#224 | Gemma 3n E2B IT | 6B (2B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#225 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#226 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#227 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#228 | Phi-3-medium | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
#229 | Yi-34B | 34B | INT4 Q8 | 370 tok/s 352 tok/s | ~25 ms ~25 ms | |
#230 | OLMo 3 7B Instruct | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#231 | Qwen2-7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#232 | Phi-3-small | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#233 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#234 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
#236 | Gemma 1 7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#237 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
#238 | ChatGLM3-6B | 6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#239 | ChatGLM2-6B | 6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
#240 | ChatGLM-6B | 6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
Unranked Models (Sorted by Speed) | ||||||
- | Agnes 3.0 Flash | 33B | INT4 Q8 | 370 tok/s 370 tok/s | ~24 ms ~24 ms | |
- | AREX-2 | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
- | ChatGLM3-6B-32K | 6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | ERNIE-4.5-21B-A3B | 21B (3B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 352 tok/s | ~9 ms ~9 ms ~9 ms | |
- | ERNIE-4.5-21B-A3B-Base | 21B (3B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 352 tok/s | ~9 ms ~9 ms ~9 ms | |
- | ERNIE-4.5-VL-28B-A3B | 28B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
- | ERNIE-4.5-VL-28B-A3B-Base | 28B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Falcon-40B | 40B | INT4 Q8 | 370 tok/s 352 tok/s | ~28 ms ~28 ms | |
- | Falcon-7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | Falcon2-11B | 11B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~13 ms ~13 ms ~13 ms | |
- | Falcon3-10B | 10B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~13 ms ~13 ms ~13 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Falcon3-7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | FrogNano 4B | 4B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
- | GLM-4-9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | GLM-4-9B-Chat | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | GLM-4-9B-Chat-1M | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | GreenMind-14B-R1 | 14B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~15 ms ~15 ms ~15 ms | |
- | Hunyuan Lite | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | Hy-Memory 30B A3B | 30B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Hy-MT2 30B A3B | 30B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
- | Hy-MT2 7B | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | Kimi-VL-A3B-Instruct | 16B (3B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | LensVLM 9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | LFM2.5 24B A2B | 24B (2.5B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 333 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | MaLLaM-3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | MiniCPM-V 4.7 35B A3B | 35B (3B active) | INT4 Q8 | 370 tok/s 370 tok/s | ~9 ms ~9 ms | |
- | Mistral-7B-v0.1 | 7.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | OLMo 3 7B Base | 7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | PPLX Decider v1 27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
- | PPLX Decider v1.1 27B | 27B | INT4 Q8 | 370 tok/s 370 tok/s | ~21 ms ~21 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~8 ms ~8 ms ~8 ms | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Sahabat-AI-Gemma2-9B | 9.2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | Sahabat-AI-Gemma2-9B-Instruct | 9.2B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | SEA-LION-7B | 7.1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | SmolLM3 3B | 3B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~9 ms ~9 ms ~9 ms | |
- | Spark X2.5 4B | 4B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~10 ms ~10 ms ~10 ms | |
- | Typhoon-2-8B | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | Yi-6B | 6B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~11 ms ~11 ms ~11 ms | |
- | Yi-9B | 9B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | Youtu Parsing Omni | 8B | INT4 Q8 FP16 | 370 tok/s 370 tok/s 370 tok/s | ~12 ms ~12 ms ~12 ms | |
- | AliceAI Foundation 80B A3B Base | 80B (3B active) | INT4 | 352 tok/s | ~9 ms | |
- | Kolibri 1 | 78B (3.5B active) | INT4 | 352 tok/s | ~10 ms | |
- | Kimi-Dev-72B | 72B | INT4 | 333 tok/s | ~44 ms | |
- | Typhoon-2-70B | 70B | INT4 | 333 tok/s | ~43 ms | |
Inference Capacity
Maximum model parameter class runnable by weight precision.
Context:
4-bit
32B Models
8-bit
32B Models
16-bit
14B Models
Fine-Tuning Capacity
Hardware capacity for local training and adapter fine-tuning.
Context Length:
QLoRA
32B Models
LoRA
14B Models
Full Parameter Training
7B-8B Models
Workload Recommendations
Guidance for optimal precision and operational ceilings on NVIDIA RTX 6000 Ada Generation (48GB).
Inference Sweet Spot
Optimized for 14B models at uncompressed or Q8 precision, and 32B models at Q4 with up to 32k context.
Bandwidth & Speed Profile
With 960 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 213 tok/s on an 8B Q4 model and 25 tok/s on a 70B Q4 model.
Frequently Asked Questions
©2025 ApX Machine Learning