PewDiePie’s Ajax AI Model Takes Aim at Big Tech

Ajax is built from a fine-tuned version of Qwen3.5-9B, giving PewDiePie’s project a much different approach from the enormous cloud-based AI systems operated by companies such as OpenAI, Google, and Anthropic.

His goal is simple but ambitious: prove that useful AI does not always need to depend on a massive remote data center. Instead, smaller models could potentially run directly on a user's own computer.

What Is PewDiePie’s Ajax AI?

Ajax is a fine-tuned AI model designed to run locally. Rather than sending every request to a cloud-based AI service, the concept focuses on performing useful tasks directly on local hardware.

According to PewDiePie, Ajax is being developed around practical computer tasks such as browsing, email and calendar management.

Ajax Is Based on Qwen3.5-9B

One of the most interesting parts of the project is its foundation. Ajax is based on a fine-tuned Qwen3.5-9B model.

Instead of attempting to build a huge AI system from scratch, the project takes an existing open model and adapts it for a more specific purpose.

This approach can make experimentation much more accessible. Developers and independent AI enthusiasts can modify smaller models without requiring the enormous computing infrastructure associated with frontier AI development.

What Can Ajax Do?

PewDiePie says Ajax has been designed to work with several everyday computer activities. The project is particularly focused on tasks that normally require interaction with different applications.

Browsing

Ajax is intended to assist with browsing-related tasks, potentially allowing the AI to interact with information on the web instead of simply answering questions from a static knowledge base.

Email

Email is another major target. A locally running assistant could potentially help users read, organize, draft or manage messages while keeping more of the workflow under their own control.

Calendar Tasks

Ajax is also being developed for calendar-related activities. This points toward PewDiePie’s broader vision of an AI assistant that can actually perform tasks instead of simply generating text.

PewDiePie Used Heretic to Change the Model

Another controversial part of the project involves Heretic, a technique or tool used to modify the behavior of AI models.

PewDiePie says he used it to remove many refusal patterns from Ajax. At the same time, he says the project still maintains limits around harmful requests.

This highlights one of the biggest differences between locally controlled AI and commercial AI services: users and developers can have considerably more control over how an open model behaves.

The OpenAI Training Controversy

The most controversial part of Ajax's development reportedly involved AI model distillation.

PewDiePie says OpenAI banned him twice after he used outputs from its models to help train Ajax through distillation.

The situation raises an important question for the AI industry: where should the line be drawn between using AI-generated outputs for development and competing directly with the model that generated those outputs?

PewDiePie criticized the restriction and compared it with the broader way large AI companies have developed their models using enormous amounts of publicly available information.

Ajax Is Already Being Tested

PewDiePie says Ajax v1 already succeeds roughly 9 out of 10 times on the tasks it was specifically built to perform.

That does not mean Ajax is a replacement for every major AI model. Instead, the reported performance is focused on the particular workflows PewDiePie has designed for it.

He is also already working on the next version, suggesting that Ajax is being treated as an ongoing experiment rather than a finished product.

Why Local AI Could Become Important

PewDiePie’s project points toward a broader trend in artificial intelligence: smaller models running locally on personal hardware.

Cloud AI has obvious advantages. Large data centers can provide enormous computing power and access to increasingly capable models. But cloud systems also mean that users depend on external infrastructure and services.

Local AI takes a different approach.

• More control over the AI environment
• Potentially greater privacy for local workflows
• Less dependence on cloud services
• Ability to customize models for specific tasks
• Potentially lower ongoing API costs

Could Smaller AI Models Replace Big Cloud Models?

Probably not for everything. Frontier models still have major advantages in complex reasoning, multimodal tasks, coding and large-scale agentic workflows.

But everyday AI tasks are different.

A user may not need the world's most powerful AI model simply to organize email, check a calendar, browse information or automate repetitive computer actions.

That is where smaller specialized models could become extremely useful.

What PewDiePie’s Ajax Project Really Represents

Ajax is more than another AI model. It represents a growing interest in personal AI — artificial intelligence that runs on hardware controlled by the user rather than entirely inside a company's cloud.

PewDiePie has already experimented with hardware, local models and AI customization. Ajax is another step toward his larger vision of making powerful AI tools available without depending entirely on Big Tech.

Whether that vision becomes mainstream remains to be seen. But the rapid progress of smaller open models suggests that local AI is becoming increasingly difficult to ignore.

Frequently Asked Questions

Final Thoughts

PewDiePie’s Ajax project shows how the AI landscape is expanding beyond massive cloud models. With smaller open models becoming increasingly capable, individuals can experiment with AI in ways that were once limited to large technology companies.

Ajax may still be an early experiment, but its philosophy is interesting: AI does not always have to live in the cloud. For many everyday tasks, a smaller model running directly on your own computer could eventually be enough.