An enthusiast launched a modern AI model on a 20-year-old Xbox 360 console

Social network user X Andrei David managed to install and run an artificial intelligence model on the Xbox 360 console based on the Llama2.c engine, which was written in the C programming language by former OpenAI and Tesla employee Andrej Karpathy. The enthusiast was prompted to do this by the work of specialists from EXO Lab, who at the end of last year launched the Llama Large Language Model (LLM) on a 26-year-old computer with Windows 98.

Image Source: Shutterstock

Although the enthusiast used the AI ​​model on the same engine as EXO Lab, he had to optimize the program code for the console’s PowerPC processor and memory management functions. The main difference is that in PowerPC, a little-endian system stores the most important values ​​first, whereas the Intel Pentium II processor used in PCs stores the smallest values ​​first. To ensure the AI ​​model worked correctly, David had to add a byte exchange system and ensure that all data created and stored was properly aligned to the 128 bytes in memory required by the Xbox 360 memory subsystem.

As a result, the enthusiast launched the AI ​​model on an Xbox 360 with a Xenon processor based on PowerPC architecture with 3 cores and an operating frequency of up to 3.2 GHz, as well as 512 MB of RAM. Running a large language model based on Llama 2 on a decades-old device is a significant achievement. However, one Platform X user noted that the 512 MB of RAM in the Xbox 360 should be enough to run Hugging Face’s SmolLm algorithms or the 4-bit 0.5B Qwen2.5 model. “Challenge accepted,” David wrote in response. This means that it will try to run other AI models on the Xbox 360 in the future.