A 125-billion-parameter model now runs on an ordinary gaming PC.
A free installer called Strata splits a huge Qwen model across your graphics card and system memory. It needs a card with 12 gigabytes of video memory and reports 53 to 94 tokens a second. The project is ten days old and has already passed 9,600 stars on GitHub.
Why it mattersA model this size used to mean renting cloud time or buying a server. Now the same work can stay on your own machine, with nothing sent out.
▶ Related video: How Strata Runs a 125B Model on a 12GB GPU (Qwen3.8-Flash-Next)
Read in the app — free, in 9 languages
Related stories
Meta open-sourced the firmware for building your own Muse gadget.2026-10-04A Linux desktop now refuses code written by AI outright.
2026-10-04DeepSeek put its agent on your desktop and opened the code.
2026-10-03Google is cutting free Gemini users down to its smallest model.
2026-10-03Meta denies its agent read a writer's messages without permission.
2026-10-01