Build a studio-grade AI stack for literally zero dollars.
Free compute, free tokens, free media generation, and the backend to ship it — researched with frontier models, reconciled against primary sources, and filtered hard on what you can actually monetize.
The guides
Evergreen, deeply-researched, and kept current.
AI & Tokens 101
Start here: what tokens are, why this is free, and how we got to a $0 stack.
Read the guide → Compute 🖥️Free Compute
Always-free CPUs & VMs, free GPU notebooks, serverless GPU, trial credits, fine-tuning.
Read the guide → LLMs 🪙Token-Maxxing
Max frontier-model usage for $0: free LLM APIs, the BYO-key move, China's cheap tokens.
Read the guide → Media 🎙️Audio Generation
Free music, voice cloning & transcription — with the commercial-licensing landmines flagged.
Read the guide → Media 🎬Media Post-Production
Finish for $0: upscale, 60fps, stems, mastering, subtitles, dubbing, 3D & matting.
Read the guide → Build 🚀Ship-It Stack
The backend to deploy an AI app for $0 — DBs, vector, embeddings, hosting, auth.
Read the guide →🛠️ Fresh open-source releases
Latest versions of the free tools we track.
Demucs v4.0.1
SOTA open stem separation (MIT)
llama.cpp b9623
CPU/Metal inference for GGUF models
vLLM v0.23.0
High-throughput open inference server
llama.cpp b9616
CPU/Metal inference for GGUF models
Ollama v0.30.8
Run open models locally, unmetered
Unsloth v0.1.463-beta
2× faster QLoRA on a free T4
Unsloth v0.1.464-beta
2× faster QLoRA on a free T4
Langfuse v3.185.0
Open LLM tracing, evals, prompt mgmt
Why does this exist?
Every year more of the AI stack falls to $0 — first tokens, then audio, then finishing, then the backend. What stays scarce is trustworthy info on which free option is actually free, actually good, and actually safe to monetize. That's this site.