Llama-3.1-70B-Instruct
meta-llama/Llama-3.1-70B-Instruct
70B Llama, requires 2x48GB / 80GB-class GPUs for full precision.
Parameters
70B
Context
128K
Downloads
3.1M
Likes
6.1K
Architecture
llama
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 169.9 GB RAM 254.8 · Disk 242.4 | 193.1 GB RAM 289.7 · Disk 268.4 | 297.5 GB RAM 446.3 · Disk 307.4 |
| FP88bit | 87.4 GB RAM 131.1 · Disk 140.7 | 106.9 GB RAM 160.4 · Disk 166.7 | 207.6 GB RAM 311.4 · Disk 205.7 |
| INT88bit | 87.4 GB RAM 131.1 · Disk 140.7 | 106.9 GB RAM 160.4 · Disk 166.7 | 207.6 GB RAM 311.4 · Disk 205.7 |
| AWQ4bit | 47.7 GB RAM 71.6 · Disk 91.8 | 65.4 GB RAM 98.1 · Disk 117.8 | 164.3 GB RAM 246.4 · Disk 156.8 |
| GPTQ4bit | 48.3 GB RAM 72.4 · Disk 92.4 | 66.0 GB RAM 99.0 · Disk 118.4 | 164.8 GB RAM 247.3 · Disk 157.4 |
| GGUF4bit | 49.3 GB RAM 73.9 · Disk 93.7 | 67.0 GB RAM 100.6 · Disk 119.7 | 166.0 GB RAM 248.9 · Disk 158.7 |
GPU Recommendations
1× H200 SXM
141 GB VRAM
≈ 141.9 tok/s
Effective production throughput ≈ 99.3 tok/s
score 0.5948
8× Tesla V100
128 GB VRAM
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.4947
8× H200 SXM
1128 GB VRAM
≈ 454.2 tok/s
Effective production throughput ≈ 317.9 tok/s
score 0.2197
1× A100 80GB
80 GB VRAM
≈ 60.3 tok/s
Effective production throughput ≈ 42.2 tok/s
score 0.6682
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
A100 80GB
80 GB VRAM
$876
/mo · vast
H100 SXM
80 GB VRAM
$1,170
/mo · vast
H200 SXM
141 GB VRAM
$2,555
/mo · lambda
B200
180 GB VRAM
$3,013
/mo · vast
B200 SXM
180 GB VRAM
—