Choosing an AI model

Check the architecture and file format

Look up your runtime’s current supported architectures and import instructions. A format label alone does not establish compatibility. GGUF packages weights and metadata, but a GGUF file is not a promise that every runtime can load that model.

Hugging Face’s GGUF format guide ↗ explains the format and file inspection. If you use Ollama, follow its separate Safetensors and GGUF import instructions ↗ for your files.

Compare versions and full memory needs

Open the original files and compare the available versions and quantization types. Check memory for the exact file, context length and hardware settings you intend to use, rather than relying on parameter count alone.

For example, LM Studio’s loading and resource-estimation guide ↗ explains estimating memory before loading, including context length and GPU offload settings. Use estimates and loading guidance for your chosen runtime; this directory does not estimate hardware fit.

Read the model card and license

Check the original model card for intended uses, limitations, base models, training details and any evaluation reports. Read the full license and base-model restrictions before your intended use; a license tag is not permission by itself.

Hugging Face’s model-card guide ↗ describes these fields and explicit base-model relationships. Follow the creator’s setup instructions, then check behavior with your own prompts. Creator-reported labels do not establish accuracy or suitability.

Browse the listings

All models · GGUF models and versions · Model families

Uncensored models · Abliterated models · Fine-tuned models

Each model’s detail page links to the original model card and files. Read how the listings and labels are collected.