Llama 3 70B API Pricing
Meta (open weights) · 8K context window · Best for original Llama 3, shorter context
Last updated April 25, 2026 · Data refreshed every 6 hours
Input
$0.60
Output
$0.60
Context
8K
Vendor
Meta (open weights)
All prices in USD per million tokens. Live as of April 25, 2026.
Cost Examples
| Workload | Tokens | Cost |
|---|---|---|
| 1M chat messages (~500 tok in / 500 tok out) | 1000M | $600.00 |
| Process 10K PDFs (~2K tok each, output 200 tok) | 22M | $13.20 |
| Generate 100K tweets (~280 tok out, 50 tok prompt) | 33M | $19.80 |
| 1 chatbot conversation per user × 10K users (~5K tok each) | 50M | $30.00 |
| Summarize 1K technical articles (~5K in / 500 out) | 6M | $3.30 |
Self-Hosting on Cloud GPUs
Because Llama 3 70B has open weights, you can self-host on cloud GPUs instead of paying API rates. Estimated cost on a single H100 spot instance:
API cost / 1M tokens
$0.60
Self-host / 1M tokens (H100 spot)
~$0.667
Self-hosting wins when your sustained throughput justifies a dedicated GPU. See H100 pricing.
Related Tools & Pages