Best Free Web Resources

Local AI Tools & Digital Engineering Resources

What can I run?

What do you want to do?

Free calculators and hardware data for running AI models on your own machine — every figure computed from published specifications, never benchmarked in-house.

See what my PC can run Pick your VRAM, get the models that fit and the quantization each one needs. Pick a GPU for AI What each VRAM tier actually unlocks, and why capacity beats raw speed. Fine-tune a model LoRA, QLoRA or full fine-tuning — the memory each one really costs. Train an image LoRA Flux, SDXL and SD 3.5, including the caching step that saves gigabytes. Compare my options GGUF vs AWQ, Ollama vs vLLM, fine-tuning vs RAG — and hardware head-to-heads.

The calculators

Answer a sizing question in seconds instead of reading a forum thread.

VRAM Calculator Model weights, KV cache and framework overhead for any model and context length. Inference GGUF Size Calculator File size at every quantization level, from bits per weight measured on real GGUF files. Quantization Fine-tuning VRAM Weights, gradients, optimizer state and activations for LoRA, QLoRA and full fine-tuning. Training LoRA Training VRAM Image-model LoRAs for Flux, SDXL, SD 3.5 and Qwen-Image at any rank and resolution. Training Image Generation VRAM Peak memory by precision and offload strategy, computed per component. Diffusion Mac VRAM Optimizer How much unified memory your Apple Silicon GPU can actually address. Apple Silicon

Start from the memory you have

Every model in our directory, scored against what a card that size realistically gives you.

8GBEntry 12GB14B models 16GBRecommended 24GB32B models 32GBLong context 48GB70B models 64GBWorkstation 128GBUnified memory

Latest guides

Hardware sizing, quantization and local-model setup, with sources on every claim.

Directories and data

The reference tables behind the calculators.

Models Directory Parameter counts, licences and memory requirements for open-weight models. 30 models Hardware Matrix Laptops, desktops and GPUs profiled against the models each one can run. 79 devices Licence Checker What each model licence actually permits, including commercial use and derivatives. Before you ship API Pricing & ROI Current per-token rates, so you can work out your own break-even against hardware. Local vs cloud

Pick a runner

The software that actually loads the model, and what each one costs in memory.

Ollama Every model’s memory cost and the pull command that fetches it. Easiest start LM Studio Memory cost, lms commands, and MLX builds on Apple Silicon. Desktop app llama.cpp The engine underneath most runners, with the flags that control memory. Full control vLLM Serving throughput, and why it reserves the whole card up front. Replacing an API Open WebUI The chat interface to put in front of a runner, and what to pair it with. Front end ComfyUI Peak VRAM for every diffusion model, by precision and offload mode. Image generation Self-Hosted AI What running it yourself protects, and the three layers of the stack. Start here for why

Free web resources

The wider BFWR archive — design assets, development tutorials and free tools.

Technology Coding Tutorials Freebies Graphics WordPress SEO