All Calculators
performance
LLM Inference Throughput Calculator
Calculate LLM inference throughput (tokens/second) for different GPUs and model sizes.
13 B params
1 GPUs
Throughput
308tokens/sec
Cost per 1M tokens
$2.53
Estimate: Serving 13B model on 1 H100 with FP16. Actual throughput depends on sequence length, batching, and framework (vLLM, TensorRT-LLM, Triton). Use vLLM with PagedAttention for best results.
Note: These calculators provide estimates based on typical configurations and pricing. Actual costs and performance may vary. Contact Harch Corp for detailed quotes and custom infrastructure planning.