Explore AI
meta-llama prices
8 models, with what one long document actually costs.
Prices checked 2026-09-20 yesterday
Buy tokens direct: meta-llama console →
“Task” = a 10,000-token document in, a 1,000-token answer out.
| Model | Context | In /M | Out /M | One task |
|---|---|---|---|---|
| Meta: Llama 3.2 1B Instruct | 60k | $0.03 | $0.20 | $0.0005 |
| Meta: Llama 3.2 3B Instruct | 131k | $0.05 | $0.33 | $0.0008 |
| Meta: Llama 3.1 8B Instruct | 131k | $0.05 | $0.08 | $0.0006 |
| Meta: Llama 4 Scout | 1311k | $0.10 | $0.30 | $0.0013 |
| Meta: Llama 3.3 70B Instruct | 131k | $0.10 | $0.32 | $0.0013 |
| Meta: Llama Guard 4 12B | 164k | $0.18 | $0.18 | $0.0020 |
| Meta: Llama 4 Maverick | 1049k | $0.19 | $0.65 | $0.0025 |
| Meta: Llama 3.1 70B Instruct | 131k | $0.40 | $0.40 | $0.0044 |
Cheapest paid model from this provider: Meta: Llama 3.2 1B Instruct at $0.03 per million input tokens.
Going direct to meta-llama usually costs slightly less than going through an aggregator, at the price of a separate account and bill. Worth it at volume; rarely worth it while experimenting.