Llama 3.3 70B Instruct: price, context window and capabilities
Llama 3.3 70B Instruct is a model from Meta (meta-llama/llama-3.3-70b-instruct). It supports tool use. It costs $0.220 per million input tokens and $0.500 per million output tokens, with a 131K-token context window.
What it costs in practice
Using the listed prices: a chat turn of 2,000 input and 500 output tokens costs about $0.0007, so 10,000 such turns cost about $6.90. Summarising a 100,000-token document into 2,000 tokens costs about $0.023. A long agent run that reads 1 million tokens and writes 50,000 costs about $0.245.
How it compares
Among the 40 models we track, Llama 3.3 70B Instruct ranks #15 by input price and #7 by output price (1 is cheapest), and #39 by context window (1 is largest). Models with the same tool and reasoning support that cost less per output token: Ministral 3 14B 2512 ($0.200), Llama 4 Scout ($0.300). Models with a larger context window at the same or lower output price: Llama 4 Scout (1.3M), GPT-6 Luna Pro (1.1M), GPT-6 Luna (1.1M), DeepSeek V4 Pro 0423 (1.0M).
More from Meta
- Llama 4 Maverick: $0.188 in / $0.652 out, 1.0M context
- Llama 4 Scout: $0.100 in / $0.300 out, 1.3M context
- Llama 3.2 1B Instruct: $0.027 in / $0.201 out, 60K context
Prices and limits come from OpenRouter's public model catalogue, last synced 4 Oct 2026. Provider pricing can differ from OpenRouter's and changes often; confirm on the provider's own page before you commit. See the guide to choosing a model or the full comparison.
Feed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.