LLM runs on Commodore 64 in impressive display of 80s tech staying power

The 42-year-old Commodore 64 just became the oldest computer capable of running a large language model, showcasing the remarkable versatility of early computing hardware in the age of AI. While modern AI companies race to optimize their models for efficiency on contemporary devices, developer Maciej Witkowiak has taken a dramatically different approach by successfully porting a simplified LLM to run on 1982 technology, demonstrating how even the most basic computing platforms can participate in today’s AI revolution.

The big picture: Developer Maciej Witkowiak has successfully ported a simplified version of Llama 2 to run on a Commodore 64 computer from 1982, creating “Llama2.c64.”

• The project runs a 260K “tinystories” model on the vintage hardware, requiring at least 2MB of RAM expansion.
• While extremely limited compared to modern AI systems, the project represents an impressive technical achievement in running AI on hardware that predated large language models by decades.

How it works: The AI has been specifically adapted to the unique constraints of the Commodore 64’s hardware environment using the oscar64 compiler.

• Rather than functioning as a chat model, the system is designed to continue stories when given a prompt, similar to “prompting a 3-year-old child with the beginning of a story.”
• The system generates text token by token, with the developer noting that “waiting for the next token on a C64 is just as exciting as waiting for one coming from DeepSeek running on your laptop.”

Behind the performance: As demonstrated in images shared by the developer, the AI produces simple, childlike stories that reflect both the model’s limited training and the hardware constraints.

• Despite the limitations, the system successfully continues narrative prompts with coherent, if simplistic, story content.
• Witkowiak lists the project’s strengths as providing complete user control, functioning on extremely modest hardware, and delivering an engaging user experience despite its technological limitations.

Why it matters: The project represents an interesting counterpoint to the industry’s focus on increasingly powerful hardware for AI applications.

• While major tech companies invest in cutting-edge chips and powerful computing clusters, this project demonstrates how even rudimentary AI capabilities can be implemented on historically significant computing platforms.
• The experiment highlights the remarkable versatility of early computer architecture while providing perspective on how far AI and computing technology have evolved over four decades.