Quick Run
Start a mini model run in under a minute and inspect loss, steps, and samples as they stream.
App
MicroLLM
On-device AI
Embedded Python
Learn by Building
MicroLLM
An on-device mini model lab for learning tokenizer, autograd, attention, and transformer assembly with embedded Python.
Runtime
Logs, metrics, and samples
Course Path
Tokenizer → Transformer
Blueprint Lessons
Learn tokenizer, autograd, attention, and transformer assembly through structured checkpoints.
Local Workspace
Fork the reference project, edit Python files, and keep project data in a local sandbox unless you export it.
About MicroLLM
A focused mini-model lab for understanding AI systems through visible code, visible logs, and visible boundaries.
Last Updated: March 13, 2026
Website: https://zelonai.com/app/microllm/
MicroLLM
MicroLLM is a mini model lab for iPhone and iPad.
It helps learners and builders run, inspect, and modify small language model workflows directly on-device, with embedded Python and structured learning paths.
Core Experience
- Quick Run: Launch a mini model and watch logs, metrics, and generated samples
- Learn by Building: Move from tokenizer and autograd to attention and transformer assembly
- Editable Projects: Fork the reference project, inspect code, and run your own experiments
- Local Workspace: Keep project files in an app sandbox and export only when you choose
Product Positioning
- Local-first: The current release focuses on on-device experiments instead of cloud execution
- Transparent runtime: Code, logs, tests, and project state are visible to the learner
- Education + prototyping: Designed for teaching demos, self-study, and quick algorithm validation
Contact
- Support Email: [email protected]
- Website: https://zelonai.com/app/microllm/
- Support Page: https://zelonai.com/app/microllm/support
© 2026 MicroLLM. All rights reserved.
Docs & Support
MicroLLM legal, support, and about pages under a single URL prefix.