LLM Tutorials

Learn how to run Qwen3.8-Flash-Next GGUF locally with llama.cpp on an RTX PRO 6000, then connect it to OpenCode for a fully local agentic coding setup.

A hands-on look at Anthropic's Fable 5 model in Claude Code: how it handles a one-shot build, its safety refusals and Opus fallback, and when to use it.

Use Unsloth Desktop for local AI workflows, including LLM inference, fine-tuning, vision models, image and video generation, OpenCode coding agents, and local APIs from one desktop application.

Learn how to run fast, accurate, multilingual speech-to-text locally on a CPU using Microsoft VibeASR.cpp, with file transcription, real-time streaming, and a Gradio web interface on Windows, Linux, and macOS.

Learn to build a document QA system in Python with PageIndex, then benchmark on a real Federal Reserve financial report it against a vector RAG baseline using FAISS.

Run the latest DeepSeek V4 Flash model on a multi-GPU setup with Unsloth Studio, connect it to OpenCode, and use a local AI coding agent to build an interactive stock analytics website.

Learn how Claude Code sub-agents can explore codebases, plan implementations, and complete multi-step development tasks with minimal supervision.

Run Gemma 4 12B QAT locally with llama.cpp and see how Quantization-Aware Training improves 4-bit GGUF inference, reduces VRAM usage, boosts speed, and makes local AI more stable on consumer GPUs.

Claude Code is not free by default. Learn about all alternative paths, with the exact setup commands and an honest account of what each one trades away from the original.

Learn how to fine-tune DiffusionGemma 26B-A4B for biomedical question answering on PubMedQA using Unsloth, LoRA, Hugging Face, and a RunPod H100 GPU.

Learn how to leverage Google's latest multimodal models to build high-performance, cost-effective image and video generation pipelines in Python.

Learn how to run Bonsai 27B, a compact 1-bit Qwen3.6-27B model, locally using llama.cpp, an OpenAI-compatible API, a web UI, and the Pi coding agent.