Open-Source AI Leaderboard Matches Local Models to Your Hardware
An open-source tool named whatmodel.sh has been launched to help users discover which local AI models their specific hardware can run by inputting RAM and GPU specs. It provides unified rankings, terminal access via curl, and direct commands to run the suggested models. It bridges the gap between hardware constraints and model performance, enabling developers and hobbyists to easily identify top-tier local AI models they can run for free. By comparing local models directly against API-based models, it highlights the practical viability of running advanced AI locally. The tool aggregates nearly 20,000 published benchmark results into a single unified score for each model, with all sources fully linked. The codebase, benchmarking scripts, and scoring logic are completely open-source on GitHub.
## BACKGROUND
Running local Large Language Models (LLMs) requires managing hardware memory limitations, which is often addressed by quantization techniques like 4-bit compression to reduce memory requirements. Benchmarking aggregation combines multiple evaluation metrics to provide a single standardized comparison framework for different models.