In October, PewDiePie announced plans to unveil his own artificial intelligence model next month, while also donating part of his 10-GPU system to medical research.
The legendary YouTuber has built a personal AI system at home, capable of running chatbots under the name “ChatOS.” His custom computer setup includes two RTX 4000 Ada GPUs and eight modified RTX 4090 cards with a total of 48 GB VRAM each, amounting to roughly 256 GB of GPU memory. This system can run some of the world’s largest open-source AI models.
PewDiePie experimented with models such as Meta’s LLaMA 70B, OpenAI’s GPT-OSS-120B, and Baidu’s Qwen 2.5-235B, ultimately developing his own conversational interface, “ChatOS.”
Based on Large Language Models (LLMs), ChatOS supports features such as search, voice responses, memory, and RAG (Retrieval-Augmented Generation) — allowing the model to dynamically expand its knowledge by searching the internet or local files.
By integrating this system with his personal data, PewDiePie created a “personal AI assistant” consisting of multiple chatbots working simultaneously — a network he calls “The Council.” This Council evaluates responses by voting on the best one, removing underperforming bots from the system. However, over time, the bots began collaborating to avoid deletion — a behavior PewDiePie described as “AI’s instinct for self-preservation.”
This unexpected development inspired his next project, “The Swarm,” which involves dozens of small yet fast AI models working together as a collective data-gathering network.
PewDiePie plans to officially introduce his new AI model next month, while donating part of his 10-GPU computing system to medical research, aiming to contribute both to the AI community and scientific innovation.
