Meta models: prices and the hardware each needs
28 models, 6 sold per token by 8 vendors. Vendors are in alphabetical order; the price is the cheapest list price among them, per million tokens.
| Model | Size | Sold per token by | From, input | Output | GPU memory |
|---|---|---|---|---|---|
| Llama 3.1 405B | 405 B | AWS, OCI | $2.40 | $2.40 | 1014.6 GB |
| Llama 3.1 70B | 70 B | AWS, Cloudflare, Snowflake | $0.72 | $0.72 | 176.4 GB |
| Llama 3.1 8B | 8 B | AWS, Cloudflare, Databricks, Snowflake | $0.045 | $0.266 | 20.1 GB |
| Llama 3.3 70B | 70 B | AWS, Azure, Cloudflare, Databricks, DeepInfra, IBM watsonx, OCI, Snowflake | $0.10 | $0.32 | 176.4 GB |
| Llama 4 Maverick 17B | 400 B | AWS, Azure, Databricks, IBM watsonx, OCI, Snowflake | $0.24 | $0.97 | 1003.9 GB |
| Llama 4 Scout 17B | 109 B | AWS, Cloudflare, DeepInfra, OCI | $0.10 | $0.30 | 271.6 GB |
| Llama-2-13b-chat-hf | 13.02 B | not sold per token | 32.5 GB | ||
| Llama-2-13b-hf | 13.02 B | not sold per token | 32.5 GB | ||
| Llama-2-70b-chat-hf | 68.98 B | not sold per token | 172.5 GB | ||
| Llama-2-70b-hf | 68.98 B | not sold per token | 172.5 GB | ||
| Llama-2-7b-chat-hf | 6.74 B | not sold per token | 16.9 GB | ||
| Llama-2-7b-hf | 6.74 B | not sold per token | 16.9 GB | ||
| Llama-3.1-405B | 405.85 B | not sold per token | 1014.6 GB | ||
| Llama-3.1-405B-FP8 | 405.85 B | not sold per token | 507.3 GB | ||
| Llama-3.1-70B | 70.55 B | not sold per token | 176.4 GB | ||
| Llama-3.1-8B (meta-llama) | 8.03 B | not sold per token | 20.1 GB | ||
| Llama-3.2-1B (meta-llama) | 1.24 B | not sold per token | 3.1 GB | ||
| Llama-3.2-1B-Instruct (meta-llama) | 1.24 B | not sold per token | 3.1 GB | ||
| Llama-3.2-3B (meta-llama) | 3.21 B | not sold per token | 8 GB | ||
| Llama-3.2-3B-Instruct (meta-llama) | 3.21 B | not sold per token | 8 GB | ||
| Llama-4-Maverick-17B-128E-Instruct-FP8 | 401.65 B | not sold per token | 502.1 GB | ||
| Llama-Guard-3-1B (meta-llama) | 1.5 B | not sold per token | 3.8 GB | ||
| Llama-Guard-3-8B | 8.03 B | not sold per token | 20.1 GB | ||
| Llama-Guard-3-8B-INT8 | 8.03 B | not sold per token | 10 GB | ||
| Meta-Llama-3-70B | 70.55 B | not sold per token | 176.4 GB | ||
| Meta-Llama-3-70B-Instruct (meta-llama) | 70.55 B | not sold per token | 176.4 GB | ||
| Meta-Llama-3-8B (meta-llama) | 8.03 B | not sold per token | 20.1 GB | ||
| Meta-Llama-3-8B-Instruct (meta-llama) | 8.03 B | not sold per token | 20.1 GB |