Product11 publishers3 min readPublished
Qualcomm unveils two new 2nm chips: Snapdragon 8 Elite Gen 6 and 8 Elite Extreme Gen 6
Qualcomm's first 2nm flagships come in two tiers. The 30-billion-parameter on-device ceiling and the 18MB of high-performance memory sit in the Extreme only, and both tiers will be sold as flagship phones.
The Product Desk · Product desk
What happened
- Qualcomm revealed the Snapdragon 8 Elite Gen 6 and 8 Elite Extreme Gen 6, its first chips built on a 2nm process, both engineered around agentic AI workloads.
- Only the Extreme tier carries 18MB of High Performance Memory and dedicated Adreno AI matrix cores, according to PCMag.
- PCMag says the Elite Extreme uses mixture-of-experts models above 30 billion parameters for agentic AI, and attaches that figure to the Extreme alone.
- Both chips share a reengineered Hexagon NPU with the new Hexagon Element Accelerator and context windows up to 32K.
- Qualcomm reaches 2nm after MediaTek, Samsung and Apple, which shipped the iPhone 18 Pro with the 2nm A20 Pro.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision A team that wanted a 30-billion-parameter model resident on the device has to choose which part it targets, or fund two paths: local inference on the Extreme and a server call on the Elite.
- constraint "Recent flagship Snapdragon" stops working as a minimum-spec line for anything memory-bound, and the test matrix at the top of the Android range doubles.
- exposure Owners of the cheaper flagship will read reviews of features their handset cannot run locally, and the support queue absorbs the difference.
- capability Frame generation with 40% power savings in heavy gaming becomes a reason to charge more for one phone than another inside the same tier.
Mixture-of-experts routing is what makes a large model plausible on a handset at all. Instead of waking a dense model for every query, the router picks only the experts the prompt needs, so one turn draws a small share of the NPU's compute and memory, according to PCMag [6]. The weights still have to live close to the accelerator, and that is where the Extreme part's extra silicon budget went. The 18MB of high-performance memory sits right next to the graphics core, and PCMag separately credits the Extreme with a shared memory bank on the Hexagon NPU that is 50% larger [3][4]. Qualcomm says putting that memory where it is buys a 12% power improvement over the previous generation [11].
If those two descriptions point at the same pool, the base figure behind the 50% is 12MB, because 18 divided by 1.5 is 12 [18]. PCMag prints no memory figure for the regular Elite.
The shared floor is wide. Both chips get the Sensing Hub with dual micro-NPUs that Qualcomm rates at 85% better performance and 20% more efficiency on AI tasks than its predecessor [8]. Both carry the X105 5G Modem-RF with FastConnect 8800 [9]. A caption model or a summariser that fits inside a 32K context and a few billion parameters should behave the same on either part. What breaks is the feature that needs the big model resident while the user waits.
Teams aim at the top tier and assume the feature is covered. Buyers take whichever flagship the carrier discounts that month. PCMag expects both parts in upcoming phones from Motorola and Samsung, plus other models sold overseas [13], and a product page that says "flagship Snapdragon" leaves the shopper guessing which side of the memory line they bought. The media split runs the same way. Adreno Neural Fusion, which pairs AI super-resolution with real-time frame generation for 40% power savings in heavy gaming, is an Extreme feature [12]. So is 8K capture at 60fps, up from the 30fps the Galaxy S26 Ultra manages on last season's 8 Elite Gen 5 [16].
The test against a roadmap is small. For each AI feature, can the team name the model that runs on the base Elite? If it can, there is one version for both tiers and the Extreme only makes it quicker. If it cannot, the feature is an Extreme feature, and it belongs in the spec matrix as one, with a server call or a disabled toggle behind it. Both answers have a price. The first costs the team the quality ceiling it was about to market; the second costs a support thread every time someone holding the cheaper flagship reads a review of the dearer one.
How often the base part has to fall back to a server depends on sustained power and thermals under a long agent run. PCMag said it still has to test how the chips handle battery life [17].
What to watch
- Whether Samsung and Motorola disclose which of the two parts is inside each phone, or market both under the same flagship label.
- Independent battery and sustained-thermal results for both parts during a long on-device agent run.
- Whether apps get a runtime way to read the memory tier or parameter ceiling so a feature can pick its own model.