YouTube star PewDiePie is stepping further into the artificial intelligence world with a new model designed to run locally on a user’s own computer.

The creator, whose real name is Felix Kjellberg, has unveiled Ajax, a fine-tuned AI model developed for his self-hosted Odysseus workspace. The project is aimed at giving users an AI assistant that can handle everyday tasks without relying entirely on cloud-based services.

Ajax Is Built for Local AI

Ajax is a fine-tuned version of Qwen 3.5 9B, a relatively small AI model designed to work with PewDiePie’s Odysseus platform.

According to the project’s official website, Ajax is intended to operate as an always-on assistant capable of handling tasks such as web searches, browsing, email and calendar functions while keeping the experience on the user’s own hardware.

That local approach is one of the project’s biggest selling points.

Instead of sending every request to a large cloud-based AI service, local models can be run directly on compatible computers. This can give users more control over their data, although performance depends heavily on the hardware being used.

PewDiePie Says OpenAI Banned Him Twice

The launch also comes with an unusual backstory involving OpenAI.

PewDiePie said in a recent video that his OpenAI account was banned twice while he was working on Ajax. According to an email shown by the YouTuber, one of the account actions involved “distillation,” a technique in which outputs from an existing AI model can be used to help develop another model.

PewDiePie said his account was restored after one ban before another suspension occurred while he continued experimenting with AI-generated training data.

The account of what happened comes from PewDiePie and the email he displayed. It does not represent an independent explanation from OpenAI about his specific case.

The AI Project Is Connected to Odysseus

Ajax is being developed as part of Odysseus, PewDiePie’s self-hosted AI workspace.

The broader project is designed to combine AI chat, agents, research tools, documents, email, notes and calendar functions in a locally controlled environment.

The approach reflects a growing interest in smaller AI systems that can perform useful tasks without requiring the enormous computing resources associated with the largest commercial models.

PewDiePie has argued that smaller models can make more sense when they are trained or tuned for specific jobs rather than trying to build one giant system capable of doing everything.

Ajax Is Not Yet Available as a Finished Public Download

Despite the announcement, people looking to download Ajax immediately may have to wait.

The project’s official website currently lists Ajax as coming soon, with a scheduled arrival shown for October 3, 2026.

That means the current announcement is more of a preview of what PewDiePie plans to release than a finished public rollout.

The project is also expected to undergo additional training, testing and benchmarking before the next version is released.

PewDiePie Wants More Control Over AI

The Ajax project reflects PewDiePie’s broader interest in giving people more control over artificial intelligence.

His approach centers on smaller models that can run on personal computers and work alongside a self-hosted software environment. That could appeal to computer enthusiasts and AI users who want more control over where their data and AI workloads are handled.

At the same time, local AI comes with tradeoffs. Running models on personal hardware can require significant computing power, and users remain responsible for securing their own systems and data.

A YouTube Star Takes a Different Route Into AI

PewDiePie is better known for his years as one of YouTube’s biggest creators than for software development, making Ajax an unusual addition to his online career.

The project puts a familiar internet personality into a rapidly expanding field dominated by major technology companies and specialized AI startups.

Whether Ajax becomes a widely used AI tool remains to be seen, but its combination of a massive creator audience, local computing and personal AI gives the project an unusual place in the crowded AI landscape.