WhichLLM is an open-source command-line tool that answers a deceptively simple question: which local LLM will actually run — and run best — on your hardware? It detects your GPU and memory, then ranks models by real, recency-aware benchmark results rather than parameter counts or hype. The project has earned about 6,800 stars on GitHub, is MIT licensed, and is under active development in 2026.
Choosing a local model is usually guesswork spread across leaderboards that ignore whether a model fits your VRAM or how fast it will generate on your specific GPU. WhichLLM automates the whole decision: run one command and it profiles your machine, filters the current model landscape to what genuinely fits, and ranks the survivors by recent benchmark performance, adjusted for quantization and expected speed on your hardware. You can then download and run the recommended model straight from the same tool, and compare candidates head-to-head before committing gigabytes of disk. Its rankings are continuously refreshed from real benchmark sources, so recommendations age with the fast-moving open-model scene instead of freezing at release day. For anyone setting up local inference — for privacy, cost or offline use — it replaces hours of forum archaeology with one command.
Free. WhichLLM is open source under the MIT license with no paid version; models you download are subject to their own licenses.
Developers, hobbyists and privacy-conscious users running LLMs locally who want the best model their hardware can handle — without manually cross-checking leaderboards, VRAM tables and quantization charts.