Taplio - Build a growth routine on LinkedIn
Mini toolsGuppyLM
GuppyLM is a minimalist educational LLM project built to make language model training understandable and accessible. The repository describes it as a vanilla 8.7M-parameter Transformer with 6 layers, 384 hidden size, 6 attention heads, a 4,096-token BPE vocabulary, and a 128-token context window. It is trained from scratch on 60,000 synthetic conversations across 60 topics, runs on a single GPU in about 5 minutes, and is small enough to run fully in the browser through a quantized ONNX model via WebAssembly. Its purpose is less frontier capability and more transparent end-to-end learning of data generation, tokenization, training, and inference.
New
Text
Gen 7
Overview
GuppyLM is a tiny open-source language model that roleplays as a fish named Guppy. It has about 8.7M parameters, is trained from scratch on 60K synthetic conversations, speaks in short lowercase sentences about tank life, and is designed as an educational project to show how a full LLM pipeline can be built, trained, and even run locally in a browser.
Tools using GuppyLM
No tools found for this model yet.
Last updated: April 7, 2026
