The AI Tokens

Meta Llama Models - Token Counter & Cost Calculator

Calculate token usage and API costs for Meta's open-source Llama models. Get accurate pricing estimates for Llama 3.1 and 3.2 models via API providers.

Available Meta Llama Models

Llama 4 Maverick

Context: 1,000,000 tokens

Input: $2/1MOutput: $6/1M

Largest Llama 4 model with 400B total parameters, optimized for multimodal understanding, multilingual tasks, and agentic systems. Requires multiple GPUs.

Llama 4 Scout

Context: 10,000,000 tokens

Input: $0.88/1MOutput: $0.88/1M

Smaller Llama 4 model with 109B total parameters, can run on single GPU. Supports text + up to 5 images input, 10M token context.

Llama 3.3 70B Instruct

Context: 131,072 tokens

Input: $0.88/1MOutput: $0.88/1M

Enhanced performance relative to Llama 3.1 70B, approaches Llama 3.1 405B performance for some applications. Text-only model.

Llama 3 70B Instruct

Context: 8,192 tokens

Input: $0.58/1MOutput: $1.75/1M

Previous generation 70B model, cost-effective option via API providers like Together AI, Groq.

Llama 3 8B Instruct

Context: 8,192 tokens

Input: $0.05/1MOutput: $0.25/1M

Smaller, faster model starting as low as $0.05/1M tokens via providers like Replicate, DeepInfra.