README.md

metadata

title: GPUburnout Models
emoji: 🔥
colorFrom: gray
colorTo: blue
sdk: gradio
sdk_version: 5.50.0
app_file: app.py
pinned: true
license: mit

GPUburnout Models — Interactive Demo

Compare language models trained from scratch across multiple seasons:

• Tiny Shakespeare (3.2M params) — Character-level, trained on Shakespeare
• GPT-2 Small (134M params) — BPE tokenizer, trained on 2.8B tokens
• Llama 1B (1.04B params) — Llama architecture, trained on 11.8B tokens for $175
• GPUburnout-2B-Chat-DPO (2.06B params) — 2B parameter chat model fine-tuned with DPO on 75K examples

Built by Jun Park | Read the blog | GitHub

Last updated: April 5, 2026