Yuqi Intelligent

GPU Memory Estimator

Estimate weights, KV cache and memory headroom.

Rough weights + KV cache + supplied headroom estimate. It does not guarantee runtime, throughput, per-device memory distribution or compatibility. Use the actual model configuration.

Parameters (billions) Weight bits Transformer layers KV heads Head dimension KV bytes per element Maximum total sequence tokens Concurrent sequences Overhead / headroom (%)

Remove passwords, keys, personal data and confidential details first. Session text is not written to a file or content database by this module. Provider retention policies must be checked separately.

Result

Complete the fields and run. Local calculations are not AI responses.