No More Wasting Time: LLM Checker Knows Which Models Run on Your Hardware
The proliferation of open-source language models has made running AI locally technically accessible to an increasingly wide audience. However, selecting the right model for one's system remains a complex task: variables such as the amount of available VRAM, the type of GPU, support for AVX instructions, and various quantization formats make it difficult to navigate without consulting technical documentation and comparison tables. It is in this context that LLM Checker comes into play, a free and open-source tool that has appeared on GitHub, designed to automate the compatibility assessment between a machine's hardware and available AI models.
Once launched, LLM Checker conducts a thorough detection of system specifications: amount of RAM and VRAM, model and computing capacity of the GPU, support for hardware acceleration backends like CUDA for NVIDIA cards, ROCm for AMD solutions, and Metal for Apple Silicon. In the absence of GPU acceleration, the software does not simply return an error; it still identifies models that can operate via CPU with acceptable performance, explicitly indicating which requirements are not met for excluded models.
The tool is developed in JavaScript and requires Node.js 16 or higher. Installation is done via npm, and functionality has been verified on native Linux, macOS, and within WSL with Ubuntu on Windows. It is also distributed as a desktop application and as an executable script, with declared support for all three major operating systems.
The internal database of LLM Checker, based on SQLite and synchronized with public repositories, includes over 200 updatable models. The software supports popular architectures such as LLaMA, Mistral, Gemma, Phi, and Qwen, managing quantized variants in GGUF and GPTQ formats. Users can manually extend the catalog and report missing models or undocumented incompatibilities through the GitHub repository.
The selection logic does not limit itself to verifying whether a model can start on the system. It evaluates compatibility across multiple dimensions: expected response quality, inference speed, suitability for intended use, and ability to handle extended context windows. Those needing a model oriented towards code generation receive different recommendations compared to those looking for a generalist assistant.
A notable feature of LLM Checker is its integration with frameworks like Ollama and LM Studio. When a compatible model is not present on the system, the tool directly provides the command to download it or the link to the official source, eliminating the need to consult external catalogs. The project author clarified in the repository that LLM Checker does not automatically download models nor modify existing configurations, and it does not install add-ons: its role is purely informational and advisory.
The software can flag any specific limitations for the supported frameworks, such as the need for particular versions of the runtime or the absence of support for certain quantizations. LLM Checker includes an optional benchmark function, disabled by default to prevent high consumption. When enabled, the test measures tokens generated per second, memory usage, and GPU load, providing a preliminary estimate of the actual performance of the selected model. During normal operation, the software can show real-time operational metrics, allowing comparisons of different models on the same machine and assessing whether a hardware upgrade would yield tangible improvements.
LLM Checker is particularly suited for development environments, research labs, and personal workstations where new releases are frequently tested, as well as for those wishing to assess the feasibility of local execution before downloading large models. The code is released under the MIT license. The project is actively developed, with release notes indicating that future versions will include a plugin system for adding new models and a module for verifying compatibility with optimized inference pipelines. The GitHub repository contains detailed installation instructions, output examples, and a section dedicated to community contributions.