A 90 million parameter conversational large language model has been successfully implemented on a 2004-era Sony PlayStation Portable (PSP). While the model is capable of generating text, its inference speed is extremely slow, averaging 0.5-0.6 tokens per second, meaning responses can take several minutes. Despite its limitations, the model can produce basic creative content and answer simple factual questions, though it is prone to hallucination. AI
IMPACT Shows the potential for running AI models on low-power, legacy devices, though with significant performance limitations.
RANK_REASON Demonstration of running a small LLM on outdated consumer hardware.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →