Llama-3.1-70B: GPU memory and hosting cost
Open weights · 70.55 billion parameters · licence llama3.1
No cloud sells this model per token: to use it you run it yourself, on GPU instances you rent.
Run it yourself
It needs 176.4 GB of GPU memory at bf16, the precision it ships in.
| Cloud | Cheapest instance that fits | GPUs | An hour | A month, around the clock |
|---|---|---|---|---|
| Alibaba Cloud | ecs.gn8v-2x.8xlarge | 2 × GPU H 96 GB | $14.38 | $10,500 |
| Google Cloud | a2-ultragpu-4g | 4 × A100 80 GB | $20.28 | $14,801 |
| AWS | p4d.24xlarge | 8 × A100 40 GB | $21.96 | $16,029 |
| Azure | Standard_ND96asr_v4 | 8 × A100 40 GB | $27.20 | $19,854 |
More from Meta
Similar size, open weights
- OpenAI gpt-oss 120B
- Qwen3.6-35B-A3B
- NVIDIA-Nemotron-3-Super-120B-A12B-BF16
- Qwen3.5-122B-A10B-NVFP4
- Qwen3-Coder-Next-FP8 (Qwen)
- Qwen-72B
List prices collected 5 October 2026, refreshed every three hours. No negotiated rates or taxes.