VRAMGlass is a hardware research and price-comparison resource for people choosing equipment to run large language models locally. Its GPU rankings bring together current listings and specifications from marketplaces, helping users compare prices, video memory (VRAM), memory bandwidth, compute performance, power use, and software ecosystems across NVIDIA, AMD, Intel, and Apple devices. The site also tracks price movements and listing counts, and offers regional marketplace views for the United States, Japan, and mainland China.

Beyond graphics cards, VRAMGlass provides prices and specifications for desktop and server DDR4 and DDR5 memory. Its model information connects hardware decisions to the memory demands of local LLMs, including model weight sizes and quantization options. Guidance explains that inference requires memory not only for model weights, but also for the context cache and runtime, with longer prompts and more simultaneous users increasing requirements. Users can explore rankings, model data, and hardware listings to assess what may fit their workloads and budgets.