NVIDIA has outlined some of the advancements it has made in bringing local AI to as many users as possible, and according to the company one of the main issues with connecting PC users with local AI is the size of the AI models themselves.

NVIDIA explained during a press briefing for CES 2026 that significant improvements have been made in reducing the size of AI models, and maintaining an appropriate level of intelligence. According to NVIDIA, small AI models that were shrunk down from its larger Cloud-based counterparts were about two years behind in intelligence levels, but just in the last six months significant improvements have been made, doubling the intelligence of these small language models.
NVIDIA told the press these smaller models are getting close to matching their Cloud counterparts, and that generally the intelligence gap is closing between the two. NVIDIA added that it believes 2026 will be the "breakout year" for AI PCs, as other metric indicators such as the first months downloads for most of its PC models have jumped 10x in a year, with the top four tools also doubling in popularity during the same period.
With the reduction in AI model sizes and retention of intelligence levels, combined with the exponentially growing popularity of AI PCs running locally, NVIDIA believes 2026 will be the year AI PCs grow quickly.










