Meta API pricing
Prices per million tokens for the 4 Meta models we track, with context windows and what a typical task costs. The cheapest is Llama 3.2 1B Instruct at $0.027 input and $0.201 output per million tokens; the largest context window is Llama 4 Scout's 1.3M tokens. Updated 25 Oct 2026.
| Model | Input / 1M | Output / 1M | Context | Tools | Reasoning |
|---|---|---|---|---|---|
| Llama 3.2 1B Instruct | $0.027 | $0.201 | 60K | — | — |
| Llama 4 Scout | $0.100 | $0.300 | 1.3M | Yes | — |
| Llama 3.3 70B Instruct | $0.100 | $0.320 | 131K | Yes | — |
| Llama 4 Maverick | $0.188 | $0.652 | 524K | Yes | — |
What Meta models cost per task
Using the listed prices, for three common jobs: a chat turn (2,000 tokens in, 500 out), summarising a 100,000-token document into 2,000 tokens, and a long agent run (1 million in, 50,000 out).
- Llama 3.2 1B Instruct: chat turn $0.0002, document $0.0031, agent run $0.037
- Llama 4 Scout: chat turn $0.0003, document $0.011, agent run $0.115
- Llama 3.3 70B Instruct: chat turn $0.0004, document $0.011, agent run $0.116
- Llama 4 Maverick: chat turn $0.0007, document $0.020, agent run $0.220
How Meta compares with other providers
Across all 40 models we track, the cheapest Meta model ranks #2 by output price. For a model-by-model comparison, see the full price table or the guide to choosing an LLM.
Other providers: Anthropic pricing · OpenAI pricing · Google pricing · DeepSeek pricing · Mistral AI pricing · Qwen (Alibaba) pricing · xAI pricing · Moonshot AI pricing · Z.ai pricing
Prices and limits come from OpenRouter's public model catalogue, last synced 25 Oct 2026. Meta's own API pricing can differ and changes often; confirm on its pricing page before you commit.
Feed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.