free-gpu

Frontier-class open models on a free Kaggle TPU v5e-8: GLM-5.3-Flash 320B MoE (~64 tok/s, our own JAX engine) and Qwen3.8-27B bf16 (~130 tok/s), 262k context, prefix caching. Works with Claude Code, Codex, opencode and pi.

Run a free 35B multimodal LLM on Kaggle GPU — accessible from any OpenAI-compatible client

a shell script is used to free nvidia gpu memory

Patches to train RVC v2 voice models for free on modern Google Colab (Python 3.12 / numpy 2.x / torch 2.x). Removes fairseq, fixes PyAV+matplotlib breakage.

Qwen-Image-2.1 Uncensored GGUF on free Kaggle T4 (ComfyUI) with a ChatGPT-style mini UI + tunnel

Headless Ollama tunneling tool for running 32B LLMs on Kaggle's dual T4 GPUs — with auth, VRAM locking, and a streaming kill switch. For personal AI research within Kaggle's usage policy.

Run a free LLM server on Google Colab/Kaggle GPUs — auto-connects opencode, pi, and Hermes Agent via a secured tunnel

Deploy LTX-2.3 uncensored text-to-video (PinkCherry v1.8 GGUF + RealESRGAN upscale) on free Google Colab T4 or Kaggle P100 — gateway, tunnel, CLI, Drive auto-upload

📚 Curated, source-cited guide to free CPU/GPU/serverless/model-API/LLM quotas in 2026 — and how to stack them for $0.

A Python SDK that turns Google Colab into a remote compute runtime. Execute local functions remotely on free GPUs with a clean Python API - no notebooks required.

Reproducible experiments for running and optimizing open-weight LLMs on free GPU compute — GGUF, llama.cpp, CUDA, quantization, speculative decoding, and inference benchmarking.

Deploy Qwen-Image-2.1 (7B DiT + 8B Qwen3-VL) on a free Google Colab T4 — detached server, Cloudflare tunnel, stdlib CLI, off-peak retry watcher with VM-recycle guard

Route GPU jobs from your terminal or your AI agent across free-tier clouds (Kaggle, Colab, Lightning, local MPS). One daemon, quota tracking, explained routing, Claude Code + Codex integration.

Fine-tune ML models for $0 — complete guide with free GPU (Kaggle T4), DistilBERT, and HuggingFace deployment. Beat GPT-4 on custom data.

Live AI council in your browser. 3 small models debate, 1 large model judges. No install, no GPU - powered by NVIDIA's free API.

A lightweight setup to run Ollama (for local LLMs like LLaMA 3, Mistral, Gemma, etc.) directly in Google Colab with free GPU/TPU support. Perfect for experimenting with open-source language models without local hardware constraints.

Every free GPU option for ML training in 2026 — Kaggle T4, Colab, Lightning AI. No credit card needed. Train DistilBERT, GPT-2, ViT for $0.