Terms & Disclaimer

quantized.uk is an independent, community-driven reference site for LLM quantization. We provide metadata indexes, calculators, benchmarks, and deployment guides to help developers run models on their own hardware.

We do not host, distribute, or sell model weights. All model files are hosted by third parties (primarily Hugging Face and individual quantizers).

quantized.uk is not affiliated with, endorsed by, or sponsored by Meta, NVIDIA, Alibaba (Qwen), Mistral AI, Google, Microsoft, Hugging Face, or any other company whose products are referenced on this site.

References to product names are for identification and descriptive purposes only.

VRAM estimates, benchmark numbers, perplexity loss figures, and speed data are approximations compiled from community sources, published benchmarks, and editorial curation. They are provided for planning purposes only.

Actual memory usage and inference speed depend on your hardware, driver version, framework build, context length, batch size, and model revision. Always verify against official model cards and your own testing before production deployment.

Hugging Face download and like counts are fetched from the public Hugging Face API at build time and may be outdated between deployments.

This site and all tools are provided "as is" without warranty of any kind, express or implied, including but not limited to accuracy, fitness for a particular purpose, or non-infringement.

We do not guarantee that any model, quant format, or configuration will work on your hardware or meet your quality requirements.

To the fullest extent permitted by applicable law, the operators of quantized.uk shall not be liable for any direct, indirect, incidental, or consequential damages arising from your use of this site, including but not limited to hardware damage, data loss, failed deployments, or reliance on calculated VRAM estimates.

You are solely responsible for decisions made based on information from this site.

Llama, Meta, Qwen, Mistral, Gemma, Phi, Granite, Codestral, Mixtral, DeepSeek, RTX, GeForce, CUDA, Apple, Mac, M1, M2, M3, Hugging Face, GGUF, llama.cpp, Ollama, vLLM, ExLlamaV2, and other names are trademarks of their respective owners.

Use of these names on quantized.uk does not imply endorsement. All trademarks remain the property of their respective holders.

Each AI model is subject to its own license (e.g. Meta Llama License, Qwen License, Apache 2.0, MIT). quantized.uk does not grant any license to use model weights.

Before downloading or deploying any model, you must read and comply with the license terms on the official Hugging Face model card. Deployment guides on this site are educational and do not override those license terms.

This site links to third-party websites including Hugging Face, GitHub, and framework documentation. We are not responsible for the content, availability, or practices of external sites.

Following external links is at your own risk.

Model metadata, benchmarks, and deployment guides on this site are curated manually and may be updated without notice.

Material data changes are noted in the homepage changelog. See the About page for how often the index is refreshed.

We may update these terms at any time. Continued use of the site after changes constitutes acceptance of the updated terms.

Material changes will be noted in the site changelog on the homepage.