Qualcomm Unveils AI‑Focused Smartphone Chips, Top Model Runs 30B MoE Locally
Qualcomm introduced two new AI‑centric smartphone chips, noting that the flagship processor is capable of running a 30 billion‑parameter mixture‑of‑experts model locally.
Qualcomm has unveiled two new smartphone processors designed with artificial intelligence at their core. The company emphasized that the higher‑end chip is powerful enough to run a 30‑billion‑parameter mixture‑of‑experts (MoE) model directly on the device, eliminating the need for cloud inference.
The announcement signals Qualcomm’s intent to bring large‑scale AI capabilities to mobile hardware, a step that could enable more sophisticated on‑device applications such as advanced language models, computer vision, and real‑time personalization.
By supporting a MoE architecture, the flagship chip can dynamically allocate compute across multiple expert sub‑networks, optimizing performance and energy efficiency for complex tasks. This approach mirrors techniques used in data‑center AI models but is now being compressed for mobile form factors.
Industry observers note that running a 30‑billion‑parameter model on a smartphone represents a significant engineering milestone, potentially reshaping how developers design AI‑driven mobile experiences.
The details were reported by TechCrunch AI on September 22, 2026.
Read also
Meta’s Muse Surpasses ChatGPT’s Early Mobile Adoption
New data from Appfigures shows Meta’s Muse outpacing ChatGPT’s early mobile performance in North America.
California Enacts New Bills to curb AI Data Center Utility Costs
Governor Gavin Newsom signed a package of seven bills aimed at preventing AI data centers from shifting utility expenses to California residents.
LLM Uncertainty Warning After Near-Miss Military AI Hallucination
A GovAI researcher warned that service members must recognize the uncertainty built into large language models after an AI hallucination nearly set off a U.S. military action.
