microsoft/mdeberta-v3-base
microsoft/mdeberta-v3-base
Parameters
277.7M
Context
512
Downloads
4.9M
Likes
231
Architecture
deberta-v2
12 layers · head 64 · hidden 768
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 3.1 GB RAM 32.0 · Disk 39.8 | 5.4 GB RAM 32.0 · Disk 65.8 | 20.2 GB RAM 32.0 · Disk 104.8 |
| FP88bit | 2.8 GB RAM 32.0 · Disk 39.4 | 5.0 GB RAM 32.0 · Disk 65.4 | 19.9 GB RAM 32.0 · Disk 104.4 |
| INT88bit | 2.8 GB RAM 32.0 · Disk 39.4 | 5.0 GB RAM 32.0 · Disk 65.4 | 19.9 GB RAM 32.0 · Disk 104.4 |
| AWQ4bit | 2.6 GB RAM 32.0 · Disk 39.2 | 4.9 GB RAM 32.0 · Disk 65.2 | 19.7 GB RAM 32.0 · Disk 104.2 |
| GPTQ4bit | 2.6 GB RAM 32.0 · Disk 39.2 | 4.9 GB RAM 32.0 · Disk 65.2 | 19.7 GB RAM 32.0 · Disk 104.2 |
| GGUF4bit | 2.6 GB RAM 32.0 · Disk 39.2 | 4.9 GB RAM 32.0 · Disk 65.2 | 19.7 GB RAM 32.0 · Disk 104.2 |
GPU Recommendations
1× RTX 5080
16 GB VRAM
≈ 7384.6 tok/s
Effective production throughput ≈ 5169.2 tok/s
score 0.5758
1× RTX 5060 Ti
8 GB VRAM
≈ 3446.2 tok/s
Effective production throughput ≈ 2412.3 tok/s
score 0.654
8× H200 SXM
1128 GB VRAM
≈ 118153.8 tok/s
Effective production throughput ≈ 82707.7 tok/s
score 0.2197
1× RTX 5070
12 GB VRAM
≈ 5169.2 tok/s
Effective production throughput ≈ 3618.4 tok/s
score 0.6015
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
RTX 5060 Ti
8 GB VRAM
$39.7
/mo · vast
RTX 4070S Ti
16 GB VRAM
$49.5
/mo · vast
RTX 4080S
16 GB VRAM
$50.0
/mo · vast
RTX 5070
12 GB VRAM
$54.1
/mo · vast
RTX 5070 Ti
16 GB VRAM
$83.1
/mo · vast