Phi-4 14B
microsoft/phi-4
Parameters
14B
Context
16K
Downloads
626.7K
Likes
2.3K
Architecture
transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 37.9 GB RAM 56.9 · Disk 79.7 | 55.2 GB RAM 82.8 · Disk 105.7 | 153.6 GB RAM 230.4 · Disk 144.7 |
| FP88bit | 21.4 GB RAM 32.2 · Disk 59.3 | 37.9 GB RAM 56.9 · Disk 85.3 | 135.6 GB RAM 203.4 · Disk 124.3 |
| INT88bit | 21.4 GB RAM 32.2 · Disk 59.3 | 37.9 GB RAM 56.9 · Disk 85.3 | 135.6 GB RAM 203.4 · Disk 124.3 |
| AWQ4bit | 13.5 GB RAM 32.0 · Disk 49.6 | 29.7 GB RAM 44.5 · Disk 75.6 | 126.9 GB RAM 190.4 · Disk 114.6 |
| GPTQ4bit | 13.6 GB RAM 32.0 · Disk 49.7 | 29.8 GB RAM 44.6 · Disk 75.7 | 127.1 GB RAM 190.6 · Disk 114.7 |
| GGUF4bit | 13.8 GB RAM 32.0 · Disk 49.9 | 30.0 GB RAM 45.0 · Disk 75.9 | 127.3 GB RAM 190.9 · Disk 114.9 |
GPU Recommendations
1× H100 SXM
80 GB VRAM
≈ 495.6 tok/s
Effective production throughput ≈ 346.9 tok/s
score 0.5884
2× Tesla V100
32 GB VRAM
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.6
8× H200 SXM
1128 GB VRAM
≈ 2272.2 tok/s
Effective production throughput ≈ 1590.5 tok/s
score 0.2197
1× RTX A6000
48 GB VRAM
≈ 113.6 tok/s
Effective production throughput ≈ 79.5 tok/s
score 0.6558
💬 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
$244
/mo · vast
L40S
48 GB VRAM
$341
/mo · vast
RTX PRO 6000 WS
48 GB VRAM
$586
/mo · vast
A100 80GB
80 GB VRAM
$876
/mo · vast