DeepSeek V2 Lite
deepseek-ai/DeepSeek-V2-Lite
Parameters
15.7B
Context
32K
Downloads
449.5K
Likes
185
Architecture
moe_transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 41.9 GB RAM 62.9 · Disk 84.6 | 59.4 GB RAM 89.1 · Disk 110.6 | 158.0 GB RAM 236.9 · Disk 149.6 |
| FP88bit | 23.5 GB RAM 35.2 · Disk 61.8 | 40.0 GB RAM 60.1 · Disk 87.8 | 137.8 GB RAM 206.7 · Disk 126.8 |
| INT88bit | 23.5 GB RAM 35.2 · Disk 61.8 | 40.0 GB RAM 60.1 · Disk 87.8 | 137.8 GB RAM 206.7 · Disk 126.8 |
| AWQ4bit | 14.6 GB RAM 32.0 · Disk 50.8 | 30.7 GB RAM 46.1 · Disk 76.8 | 128.1 GB RAM 192.1 · Disk 115.8 |
| GPTQ4bit | 14.7 GB RAM 32.0 · Disk 51.0 | 30.9 GB RAM 46.3 · Disk 77.0 | 128.2 GB RAM 192.3 · Disk 116.0 |
| GGUF4bit | 14.9 GB RAM 32.0 · Disk 51.3 | 31.1 GB RAM 46.6 · Disk 77.3 | 128.5 GB RAM 192.7 · Disk 116.3 |
GPU Recommendations
1× H100 SXM
80 GB VRAM
≈ 441.4 tok/s
Effective production throughput ≈ 309 tok/s
score 0.592
2× Tesla V100
32 GB VRAM
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.5933
8× H200 SXM
1128 GB VRAM
≈ 2023.7 tok/s
Effective production throughput ≈ 1416.6 tok/s
score 0.2197
1× RTX A6000
48 GB VRAM
≈ 101.2 tok/s
Effective production throughput ≈ 70.8 tok/s
score 0.6618
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
RTX A6000
48 GB VRAM
$210
/mo · vast
RTX 5090
32 GB VRAM
$236
/mo · vast
L40S
48 GB VRAM
$341
/mo · vast
RTX PRO 6000 WS
48 GB VRAM
$482
/mo · vast
RTX PRO 6000 S
48 GB VRAM
$633
/mo · vast