hotai.news
← Back to AI Newsflash
AI News / Event details
IT Home · Published

Huawei's Kirin 9050 series chips officially adopt the new Da Vinci architecture NPU, marking the industry's first 30B MoE large model running on-device.

Automated summary · Based on collected summaries; not individually verified
Prepared by HotAI

What happened

Huawei Kirin's official WeChat account published a detailed introduction of its first 'logical folding' technology τ chip—the new Da Vinci architecture NPU in the Kirin 9050 series. It supports on-device operation of 30B large models without needing internet or waiting for cloud processing.

Background

What is a mixture-of-experts model? Expert routing and parameter counts

A mixture-of-experts model uses routing to select a subset of expert subnetworks for an input. Sparse MoE can increase total capacity while limiting active computation, but active parameter counts alone do not establish memory needs or speed. Experts are internal networks, not separate chatbots or necessarily recognizable human specialties.

Reading notes

The event summary comes from collected excerpts. Background passages quote existing HotAI articles to explain terms and do not establish new facts about this event. Check the original sources for product results, conditions and subsequent updates.

Original source

Sources and collection timeline

SourceReport titleOriginal publication timeCollected by HotAI
IT HomeHuawei's Kirin 9050 series chips officially adopt the new Da Vinci architecture NPU, marking the industry's first 30B MoE large model running on-device.2026-11-11 11:442026-11-11 11:52