Nano LM Studio is a 5 stage workbench that allows you to experiment with creating your own small language models based on your own source materials. It lets you import and clean your corpus, tokenize and configure Transformer training with CUDA acceleration, auto checkpointing, live loss and overfitting-gap monitoring, model and VRAM planning, interactive chat and attention visualisation. This is all done through a single cohesive interface that lets you see how your model is developing, where you can understand and control the whole process, and it’s completely private, and completely yours.
Released 3 September 2026 open-source under MIT license.
Top comments (0)