Running DeepSeek V4.1 locally on a 2021 gaming GPU via Strata

Strata is a free, open-source application that allows users to run large AI models locally by streaming memory overflow directly from an SSD. The app successfully runs DeepSeek V4.1 on a 2021 gaming graphics card with 46GB of RAM and an RTX 3060, achieving a writing speed of 9 to 11 tokens per second. Additionally, smaller computers can use the same application to run Qwen3.8. Open article →