Meta Llama API pricing
As of October 22, 2026, the Meta Llama API prices run from $0.08 to $1.00 per million output tokens across 8 models, as listed on OpenRouter. Llama 3.1 8B Instruct is the cheapest paid model at $0.05 input and $0.08 output per million tokens, and Llama 4 Maverick the most expensive at $0.35 input and $1.00 output.
None of the 8 models from Meta Llama has a benchmark score yet, so none is ranked. Cost per solved task is known for 2 of 8 models.
| Model | Input $/Mtok | Output $/Mtok | Context | Evidence | Index | Cost per solved task |
|---|---|---|---|---|---|---|
| Llama Guard 4 12B | $0.18 | $0.18 | 164K | Not benchmarked | No run cost | |
| Llama 4 Maverick | $0.35 | $1.00 | 524K | Not benchmarked | $1.491 | |
| Llama 4 Scout | $0.175 | $0.50 | 1.31M | Not benchmarked | $1.276 | |
| Llama 3.3 70B Instruct | $0.45 | $0.72 | 131K | Not benchmarked | No run cost | |
| Llama 3.2 1B Instruct | $0.027 | $0.201 | 60K | Not benchmarked | No run cost | |
| Llama 3.2 3B Instruct | $0.0505 | $0.3325 | 131K | Not benchmarked | No run cost | |
| Llama 3.1 70B Instruct | $0.56 | $0.56 | 131K | Not benchmarked | No run cost | |
| Llama 3.1 8B Instruct | $0.05 | $0.08 | 131K | Not benchmarked | No run cost |
Price changes
No change in a maker's own API price recorded yet. We read prices daily from OpenRouter (since October 15, 2026) and list a change only when the maker's own price moves between two of our snapshots.