Local LLM RAM and VRAM calculator
Choose your computer, model, and conversation length. The calculator estimates how much memory the model and app will need, including room for other programs, so you can decide what to download.
How much RAM or VRAM does a local LLM need?
Start with the model weights, then add memory for the conversation, the runtime, and working buffers. Keep room for your operating system and other applications. A model download that fits on your disk can still be too large for the memory available when you load it.
How to use this calculator
- Choose your computer and enter its system RAM. For a dedicated GPU, enter its VRAM separately.
- Choose the model and quantization, which controls how compactly its weights are stored.
- Set the conversation length and number of concurrent users to match your intended workflow.
- Compare the memory breakdown with the available memory. Reduce the model size or context length when there is too little room left.
What the memory calculation includes
Model-weight memory uses the recorded download size for the selected quantization when one is available. Otherwise, it is calculated from parameter count, bits per weight, and format overhead. Conversation memory is calculated separately from the model architecture, context length, cache precision, and concurrent users. Runtime overhead, working buffers, and a safety margin are then added.
RAM, VRAM, and Apple unified memory
On an Apple Silicon Mac, the CPU and GPU share the same unified-memory pool. On a PC with a dedicated graphics card, system RAM and GPU VRAM are separate constraints. Do not simply add them together and assume the entire model can run on the GPU. Moving work to the CPU changes the setup and can make responses slower.
Why conversation length changes the result
The model needs memory for the tokens it is processing as well as its weights. Longer conversations and more simultaneous users increase that allocation. For a first run, use a shorter context and one user, then increase the workload after checking the remaining memory.
Is 16GB enough to run a local LLM?
The useful answer depends on the model, quantization, conversation length, and memory pool. Enter your actual configuration rather than choosing by RAM alone. The 16GB setup guide helps you choose an initial app and model size; the calculator lets you check that choice.
For the fuller explanation, read the local LLM RAM and VRAM requirements guide. When an installed model fails to load, follow the out-of-memory troubleshooting steps.