Mohamed Meguedmi
Creator of BestLLMfor and BestLLMFor.com
Mohamed Meguedmi is the creator and author of BestLLMfor, the configurator and directory of open-weight LLMs that can run locally, as well as BestLLMFor.com, its English-language equivalent.
Convinced that generative AI can run truly at home — with no cloud API, no subscription, and no data leaks —, it has documented open-weight models since 2026 (Llama, Mistral, Qwen, Gemma, DeepSeek…), their actual VRAM usage by quantization, and their speed in tokens per second.
How the guides are tested
The guides and rankings published on this site are based on a local test fleet : a Linux workstation (CachyOS) equipped with a GeForce RTX 50-series card that runs Ollama daily, and a Mac Apple Silicon for unified memory. The installation commands, VRAM requirements, and behaviors described in the guides are verified on these machines before publication.
The complete methodology (evaluation criteria, VRAM formulas, sources) is documented on the page Methodology, and the project itself is presented on the page About.