⣿⣿⣿⣿⣿⣷⣿⣿⣿⡅⡹⢿⠆⠙⠋⠉⠻⠿⣿⣿⣿⣿⣿⣿⣮⠻⣦⡙⢷⡑⠘⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⣌⠡⠌⠂⣙⠻⣛⠻⠷⠐⠈⠛⢱⣮⣷⣽⣿
⣿⣿⣿⣿⡇⢿⢹⣿⣶⠐⠁⠀⣀⣠⣤⠄⠀⠀⠈⠙⠻⣿⣿⣿⣦⣵⣌⠻⣷⢝⠦⠚⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⢟⣻⣿⣊⡃⠀⣙⠿⣿⣿⣿⣎⢮⡀⢮⣽⣿⣿
⢿⣿⣿⣿⣧⡸⡎⡛⡩⠖⠀⣴⣿⣿⣿⠀⠀⠀⠀⠸⠇⠀⠙⢿⣿⣿⣿⣷⣌⢷⣑⢷⣄⠻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⣫⠶⠛⠉⠀⠁⠀⠈⠈⠀⠠⠜⠻⣿⣆⢿⣼⣿⣿⣿
⢐⣿⣿⣿⣿⣧⢧⣧⢻⣦⢀⣹⣿⣿⣿⣇⠀⠄⠀⠀⠀⡀⠀⠈⢻⣿⣿⣿⣿⣷⣝⢦⡹⠷⡙⢿⣿⣿⣿⣿⣿⣿⣿⣿⠈⠁⠀⠀⠀⠁⠀⠀⠀⠱⣶⣄⡀⠀⠈⠛⠜⣿⣿⣿⣿
⠀⠊⢫⣿⣏⣿⡌⣼⣄⢫⡌⣿⣿⣿⣿⣿⣦⡈⠲⣄⣤⣤⡡⢀⣠⣿⣿⣿⣿⣿⣿⣷⣼⣍⢬⣦⡙⣿⣿⣿⣿⣿⣯⢁⡄⠀⡀⡀⠀⠄⢈⣠⢪⠀⣿⣿⣿⣦⠀⢉⢂⠹⡿⣿⣿
⠀⠀⠄⢹⢃⢻⣟⠙⣿⣦⠱⢻⣿⣿⣿⣿⣿⣿⣷⣬⣍⣭⣥⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣶⡙⢿⣼⡿⣿⣿⣿⣿⣿⣷⣄⠘⣱⢦⣤⡴⡿⢈⣼⣿⣿⣿⣇⣴⣶⣮⣅⢻⣿⡏
⠀⠀⠈⠹⣇⢡⢿⡆⠻⣿⣷⠀⢻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⣍⡻⣿⣟⣻⣿⣿⣿⣿⣷⣦⣥⣬⣤⣴⣾⣿⣿⣿⣿⣷⣿⣿⣿⣿⣷⡜⠃
⠀⠀⠀⢀⣘⠈⢂⠃⣧⡹⣿⣷⡄⠙⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣮⣅⡙⢿⣟⠿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠋⡕⠂
⠀⠀⠀⠀⠀⠀⠛⢷⣜⢷⡌⠻⣿⣿⣦⣝⣻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣯⣹⣷⣦⣹⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠿⠉⠃⠀
I am dedicated to pushing the frontier of AI through the development of compact, high-performance language models and custom architectures.
AgmGPT-3 represents the pinnacle of my custom architecture research. It is a 4.71M parameter model built on a full GPT architecture with numerous algorithmic improvements. It features a smart tokenizer and achieves state-of-the-art performance for its 4.71M params size class.
| Model | Architecture | Focus |
|---|---|---|
| AgmGPT-3 | Full GPT (Custom) | Current Flagship: 4.71M params, optimized transformer architecture, and high coherence. |
| AggmGPT-2 | LLM Fine-tune | Exploration of fine-tuning foundational models like Qwen and Llama. |
| AggmGPT-1.5 | Hybrid N-Gram + Attention | Scaling attention mechanisms over complex n-gram structures. |
| AggmGPT | N-Gram + Attention | Initial research into mixing modern attention with statistical n-grams. |