# Local AI Models **Entity class:** Deployment architecture **Domain:** Artificial intelligence / science / political economy **Maturity:** Developed ## Definition **Local AI Models** are systems whose weights or execution can operate on user-controlled, organizational, or community hardware rather than depending entirely on a remote provider. ## Mechanism and Consequence Local execution changes the politics of control: capability can continue diffusing despite centralized policy, while privacy, ownership, customization, and compute constraints move closer to the user. ## Relationships - [[wiki/Local AI Ownership|Local AI Ownership]] - [[wiki/Compute Sovereignty|Compute Sovereignty]] - [[wiki/Personal AI Agent|Personal AI Agent]] - [[wiki/AI Access|AI Access]] ## Source Dossiers - [[research/AI Is Now So Good They Will Pay You to Quit|AI Is Now So Good They Will Pay You to Quit]] - [[research/Super Intelligence Week Index of Statements from the White House Luncheon|Super Intelligence Week Index of Statements from the White House Luncheon]] - [[research/Jensen Pushes Back on Doomers Moonshots Live|Jensen Pushes Back on Doomers Moonshots Live]] ## Simple Reminders, Quotations, and Thoughts > "The good news is that the Xi–Trump summit and Bernie Sanders’ proposal are completely irrelevant. Once local model weights are available and people can run their own models, the world will move ahead regardless of what any single person does. Humanity has always trundled forward solving problems. We will likely solve cancer in the next two years; that is an unbelievably optimistic thing to aim for. Why would anyone want to slow it down?" > **— Salim Ismail**, *MOONSHOTS Live, September 25, 2026* [[reminders/Abundance/Local AI Models Will Move Humanity Forward Despite Governments by Salim Ismail|Local AI Models Will Move Humanity Forward Despite Governments by Salim Ismail]] > "There is a model from the Chinese [company] Alibaba called Qwen 27B. It’s a 27-billion-parameter model. … It’s like 60 GB. You need 60 GB of RAM to run it by default. They got it down to like 25. A team called PrismML got it down to six. … Same model." > **— Emad Mostaque**, *Impact Theory with Tom Bilyeu, 2026* [[reminders/Local AI/The Same AI Model Shrunk From Sixty Gigabytes to Six by Emad Mostaque|The Same AI Model Shrunk From Sixty Gigabytes to Six by Emad Mostaque]] > The compressed Qwen 27B model matches the frontier Opus model that you were using in January, and it runs on 6 GB — which is a Raspberry Pi, which uses as much energy as your brain. > **— Adapted from Emad Mostaque**, *Impact Theory with Tom Bilyeu, 2026* [[reminders/Local AI/Frontier AI Now Runs on the Energy Budget of a Human Brain by Emad Mostaque|Frontier AI Now Runs on the Energy Budget of a Human Brain by Emad Mostaque]] > January’s frontier Opus model was current less than a year ago. That’s the model that was bankrupting people. That’s the model Amazon spent $500 million on, and now a model of comparable capability runs on a freaking smartwatch. > **— Adapted from Emad Mostaque**, *Impact Theory with Tom Bilyeu, 2026* [[reminders/Local AI/Yesterday's Half-Billion-Dollar AI Capability Now Runs on a Smartwatch by Emad Mostaque|Yesterday's Half-Billion-Dollar AI Capability Now Runs on a Smartwatch by Emad Mostaque]] > Recursive self-improvement can begin with a local model changing its own software harness on an ordinary computer. Fully enforcing a ban would therefore require surveillance of what people say to private models on their own machines—a technological thought police more dangerous than the capability it claims to suppress. > **— Adapted from Richard Socher**, *MOONSHOTS Live, October 2026* [[reminders/Surveillance/Banning Recursive AI Would Require a Thought Police by Richard Socher|Banning Recursive AI Would Require a Thought Police by Richard Socher]]