whatmodel is an open-source website and command-line tool that helps users determine which AI models their own hardware can run locally. Users enter their available RAM and GPU, then receive a list of compatible models, each with a ranking and a command for running it. The tool also places locally runnable models and API models on the same score for comparison. Its creator says that a 32GB Mac can now run a model scoring on par with GPT-5 entirely locally and for free, citing Qwen 3.8 27B at 4-bit as an example. Each model score is fitted from 18,731 published benchmark results, with links to the underlying sources. The project publishes its methodology, data, scoring code, and benchmark script as open source. Contributors can collect real-world tokens-per-second measurements or submit models and benchmark results with their sources through GitHub.
AI News
The latest AI releases, research, products, and industry updates.
Loading...