TCO calculator
Set a target interactivity (tokens/sec/user) and compare the throughput and cost across all chips. Values are interpolated from real benchmark data.
Llama 3.3 70B throughput & TCO calculator
The Llama 3.3 70B throughput and total cost of ownership (TCO) calculator uses measured benchmark data to estimate cost per million tokens and performance per dollar for serving Llama 3.3 70B on each chip config.