Local LLM Hardware & VRAM Calculator

About

Find which local LLMs your Windows PC or Apple Silicon Mac can run before downloading a model.

Local LLM Hardware & VRAM Calculator detects the browser-visible GPU or Apple chip and physical memory, then ranks open-weight AI models that fit the selected VRAM, context window and quantization. Hardware detection and calculations stay on your computer. Raw scan data is not uploaded to Yowox.

WHAT DOES THE LOCAL LLM CALCULATOR CHECK?

- Estimates total memory from GGUF model weights, KV cache and runtime overhead.
- Ranks compatible Llama, Qwen, DeepSeek, Mistral, Gemma and other open-weight models.
- Recalculates the shortlist when you change context length or model compression.
- Estimates generation speed for catalogued devices. Custom GPUs show Unavailable instead of a guessed speed.
- Ships with an offline catalogue of 109 GPU/chip profiles and 116 model entries in version 0.1.0.

HOW PRIVATE IS HARDWARE DETECTION?

- Detection and model-fit calculations run inside the extension.
- Raw CPU, RAM and graphics-adapter scan data is not sent to Yowox.
- The extension requests no permission to read page content, tab URLs or browsing history.
- Public catalogue updates are requested without device specifications or a unique user identifier.
- Selected GPU, memory and settings leave the extension