Llama-2-70b-hf: GPU memory and hosting cost
Open weights · 68.98 billion parameters · licence llama2
No cloud sells this model per token: to use it you run it yourself, on GPU instances you rent.
Run it yourself
It needs 172.5 GB of GPU memory at fp16, the precision it ships in.
| Cloud | Cheapest instance that fits | GPUs | An hour | A month, around the clock |
|---|---|---|---|---|
| Core42 | mi300x.1x | 1 × MI300X 192 GB | $3.50 | $2,555 |
| DeepInfra | b200.1x | 1 × B200 180 GB | $4.89 | $3,570 |
| Nebius | gpu-b200-sxm-a.1gpu-20vcpu-224gb | 1 × B200 180 GB | $8.50 | $6,205 |
| Fireworks AI | b200.1x | 1 × B200 180 GB | $13.00 | $9,490 |
| Alibaba Cloud | ecs.gn8v-2x.8xlarge | 2 × GPU H 96 GB | $14.38 | $10,500 |
| Crusoe | a100-sxm.8x | 8 × A100 80 GB | $18.40 | $13,432 |
| 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 |
| Together AI | hgx-h100.8gpu | 8 × H100 80 GB | $31.92 | $23,302 |
| OCI | BM.GPU.A100-v2.8 | 8 × A100 80 GB | $32.00 | $23,360 |
More from Meta
Similar size, open weights
- OpenAI gpt-oss 120B
- Mistral Small 4
- Qwen3.6-35B-A3B-NVFP4
- Qwen3.6-35B-A3B-FP8
- Qwen3.6-35B-A3B
- GLM-4.5-Air
List prices collected 4 November 2026, refreshed every three hours. No negotiated rates or taxes.