Manz GPU

IA Models

Here you can find AI Models for local use (text, image, audio, video...), as well as their cloud versions.

HiDream AI

Vivago

1. Models (parameters count)

Text
Image
Vision
Speech
Audio
Video
Omni

Model parameters are the learned weights inside an AI model that determine how it processes data and produces outputs. More parameters usually mean higher capacity and better performance, but also higher memory and compute requirements. Fewer parameters make the model faster and lighter, but potentially less accurate or capable.

Families (3)

Dense model
MoE
Extended model

2. Selected Model

Selected AI models from the previous graph will appear here.

3. Reducing size (quantization)

BAD q1 q2 q3 q4 q5 q6 q8 fp PERFECT

Quantization is a technique used to reduce the size of an AI model by lowering the precision of its weights and computations (fewer bits per parameter). This makes the model much smaller, faster, and easier to run on limited hardware, but it comes with a trade-off: reduced numerical precision can slightly degrade accuracy and overall performance compared to the original full-precision model.

Your hardware resources (GPU, GPU Model KV-cache, RAM):