llmspeed

llmspeed is a free browser tool for exploring how a local language model might feel on different hardware. It simulates streaming text and the wait before the first token, offers side-by-side races, and estimates whether the selected model and quantization fit in memory or require partial offload.

The project labels estimates as measured, derived or projected. Its English README describes community benchmark inputs and fallback calculations based on bandwidth, compute and memory overhead. These are estimates for a single device, not inference performed on the visitor’s computer or guaranteed purchase recommendations. The repository is MIT licensed and predates its October 17 Show HN submission.