Qualcomm unveiled two 2 nm flagship mobile SoCs on September 23, 2026 — the Snapdragon 8 Gen 6 Ultra and the Snapdragon 8 Gen 6 Super Ultra — both built around agent scheduling and on-device large-model execution, with the Super Ultra positioned as Qualcomm's most powerful mobile platform to date.

At the launch event, CEO Cristiano Amon stated that "a new moment of transformation for mobile technology has arrived, and this time it is agent-centric," framing the release as the start of an agent-first cycle rather than a future roadmap.

What does the 2 nm process change for the CPU?

Both chips pair the 2 nm node with the new Oryon CPU, Adreno GPU and Hexagon NPU. The Super Ultra's Oryon core reaches a maximum main frequency of 5 GHz. Qualcomm describes it as "the world's fastest mobile CPU," citing the following generational gains:

• Performance up 13%
• Energy efficiency up 37%
• Response speed up 15%

The headline CPU addition is Oryon FlexCache, a cache architecture that converts previously fixed per-core cache into a dynamically allocated pool. When a super core absorbs a sudden high-load task, it can draw on most of the cache pool on demand while keeping latency low.

How does the new Adreno GPU handle AI workloads?

Qualcomm rates the new Adreno GPU at +44% performance and +40% energy efficiency, calling it the "biggest generational leap ever" for the Adreno line. The GPU now integrates dedicated Matrix Cores for AI, enabling neural-network compute to run inside the graphics rendering pipeline rather than offloading to other units.

On top of that silicon, Qualcomm introduced Adreno Neural Fusion, which fuses neural processing, super-resolution and frame generation with traditional rendering. Neural Fusion is the first AI super-resolution feature to ship in mainstream engines Unity and Unreal Engine. Phones based on the Super Ultra can now capture 8K video at 60 fps.

What does the Hexagon NPU add for large models?

The new Hexagon NPU adds an Element Accelerator unit designed specifically for Transformer workloads used by generative AI and agents. It works alongside the existing vector and scalar units.

Key NPU specifications:

• Large Shared Memory capacity expanded by 50%
• INT4 prefill speed up 50%
• Precision support: INT2, INT4, INT8, FP8, FP16

Larger on-chip memory keeps context and KV-cache inside the NPU, reducing DDR accesses and freeing bandwidth for concurrent agent tasks, tool calls and longer context windows.

Which models and OEMs are on board at launch?

Qualcomm has validated complete on-device deployment of a 30-billion-parameter MoE model in partnership with Step, Wulianhuo and Longsys. Companion devices showcased at the event include the overseas debut of the QwenBook Qwen tablet and a Doubao phone.

First-wave smartphone OEMs adopting the Snapdragon 8 Gen 6 Super Ultra and Snapdragon 8 Gen 6 Ultra:

• Xiaomi
• vivo
• REDMAGIC
• REDMI
• OPPO
• OnePlus
• Motorola
• iQOO
• Honor

Xiaomi brought its upcoming 18 Pro to the stage as the first confirmed vehicle for the new platform.

What comes after the phone launch?

Qualcomm has scheduled additional Personal AI product announcements for the following two days, covering wearables and PCs. The agent-centric positioning extends beyond the smartphone silicon unveiled today into the broader Personal AI roadmap the company plans to lay out this week.