KV Cache Calculator

⚙️ Configuration

Model Type

Standard Models

Hybrid Models

DeepSeek V4 series requires Hybrid Models tab

Model Source

Preset Models

Custom Model

Select Model

Data Type

Number of Tokens

Batch Size

Tensor Parallelism (TP)

Data Parallelism (DP)

Maximum Tokens Calculator

Single-GPU Memory for KV Cache (GiB)

Memory available for KV cache (excluding model weights)

Select Hybrid Model

Deployment Method

Number of Tokens

Batch Size

Tensor Parallelism (TP)

Data Parallelism (DP)

Maximum Tokens Calculator

Single-GPU Memory for KV Cache (GiB)

Memory available for KV cache (excluding model weights)

📊 Results

📊

Configure your model and click calculate to see results.

📚 Current Model Formula Reference

📊

Select a model to see its KV Cache formula.