Hugging Face's ML Intern Turns Prompts Into Trained Models: 6 Models for ~$100
Gradio · x · 2026-10-10
Gradio (Hugging Face) launched ML Intern, an agent that turns a prompt into a trained, evaluated model published on the Hub: it plans the work, asks for a budget before spending, smoke-tests first, then trains, evaluates and ships on HF hardware. Six models were trained in a week for $100 of compute, with all prompts, runs and bills published. One example: distilling Qwen-Image 2.1's 9B prompt rewriter into a 0.8B model that runs on a CPU, with 99.7% valid outputs, 1/4 the teacher's token usage, and just $16 total compute including labeling 8,797 examples.
Related event: Hugging Face's ML Intern Turns One Prompt into Trained Models for ~$100(3 posts)→
More from coding & agent
- OpenAI launches composer predictions in Codex beta for Pro users — gaganghotra_ · 2026-10-10
- Claude Code Projects opens to all Pro and Max users off the waitlist — lydiahallie · 2026-10-10
- Ant Group's Mara Chain turns rejected candidates into stepping stones for AI auto-evolution — antgroup · 2026-10-10
- Consulting Firms Push Cookie-Cutter 'Agentic SDLC' Frameworks, Vet Slams the Playbook — DavidLinthicum · 2026-10-10
- ChatGPT Patched Moonlight: Crisp 1440p Streaming on Mac at Last — Angaisb_ · 2026-10-10
- Claude Code Projects docs detail parallel cloud threads, Remote Control and rollout limits — ClaudeDevs · 2026-10-10