MAXX2.0
A Czech language model built from scratch. It chats with you, calculates and looks facts up in the Czech Wikipedia.
We build models from scratch and make room for experiments. Right now, we’re working with language and music. There are plenty of other ideas we want to try.
APPMAXX:AI is our Czech lab. We build models from scratch, try different approaches and learn from what works and what doesn’t.
See our models01 / 05
What could be built differently? Right now, language and music. Plenty of other ideas are waiting their turn.
02 / 05
We choose and prepare our own data. MAXX 2.0 learned from Czech alone: the web, Wikipedia, Wikisource and educational PDFs.
03 / 05
No borrowed weights. We put the tokenization, architecture and training together ourselves, step by step.
04 / 05
We test what the model can do and where it fails. What doesn’t work teaches us just as much.
05 / 05
What we learn goes into the next version. That’s why MAXX and Striker both carry a 2.0.
We’re a Czech lab. MAXX learned from Czech text only, and Czech shapes every step.
Language and music come first. More experiments will follow; some may become new models, others may lead somewhere different.
Details, samples and technical notes appear here during development, not only at the end.
Two models built from scratch. One speaks Czech, the other listens to music. More experiments will follow.
A Czech language model built from scratch. It chats with you, calculates and looks facts up in the Czech Wikipedia.
One neural network that listens to a whole song and finds its drum hits, beats and bass.
If a project caught your eye, or you have an experiment in mind, drop us a line. We’d be glad to hear about it.