Privacy Policy & Terms of Use

1. Client-Side Execution & Zero Telemetry

GPUCalc operates as a fully client-side single-page architecture and static web application. All calculations, parameter configurations, slider inputs, and model memory estimations are evaluated directly inside your local web browser session using JavaScript.

We do not transmit, collect, or store your custom model parameter counts, prompt sequence lengths, or training batch configurations on any external server.

2. Local Storage Usage

The only data stored on your local browser is your visual color theme preference (localStorage.getItem('theme')) to ensure consistent dark or light styling across page transitions without screen flashing.

3. Cloud GPU Pricing & Third-Party Outbound Links

The pricing information presented in the Live GPU Matrix is compiled from public provider documentation and marketplace spot bids. When you click "Rent" or "Deploy", you are redirected to third-party cloud platforms (e.g. RunPod, Lambda Labs, Vast.ai, AWS, Google Cloud).

We do not control the terms, pricing fluctuations, or availability guarantees of these external providers.

4. Disclaimer of Warranty

GPU calculations are mathematical estimates intended for planning and sizing purposes. Real-world CUDA memory consumption may vary slightly depending on PyTorch allocator fragmentation, CUDA driver version, kernel buffer allocations, and specific model quantization implementations.

5. Contact Us

If you have questions about this policy, please reach out via our Contact & Feedback page.