Ask most people what 5G gave them and the answer is "faster downloads". That is a fair description of what has actually shipped in a lot of markets, and it is also why the technology has a reputation for over-promising. The features that made 5G genuinely different from LTE — deterministic latency, network slicing, massive device density — are not properties of the radio. They live in the core. And in a non-standalone deployment, there is no 5G core at all.
This distinction matters commercially, not just architecturally. If you are planning a product around a capability 5G was supposed to deliver, the first question is which flavour of 5G your target market actually has.
Non-standalone: a new radio on an old core
Non-standalone (NSA) was designed to get 5G to market quickly, and at that it succeeded. An operator deploys 5G New Radio for the data path but keeps the existing LTE core and LTE anchor for control signalling. Devices attach to LTE first and add 5G as a secondary carrier.
The result is a real improvement — more spectrum, higher throughput, better spectral efficiency — delivered without replacing the core network. It is a sensible engineering compromise and it is what most subscribers are connected to today.
What it cannot do is anything that depends on the 5G core's service-based architecture. There is no network slicing, no user-plane function you can place at the edge, no ultra-reliable low-latency service class. Latency improves somewhat because the radio is better, but it is still governed by an LTE control plane and a centralised core.
Standalone: where the interesting features live
Standalone (SA) replaces the core with the 5G service-based architecture. Once that is in place, three things become possible that were not before.
Network slicing. The physical network can carry logically isolated slices with their own performance characteristics and service guarantees. A hospital's critical telemetry and a stadium's video uplink can share infrastructure without either being able to starve the other.
Edge compute with a local breakout. The user-plane function can be placed close to the user, which is what allows traffic to leave the network near where it entered instead of traversing to a central gateway. This is the mechanism behind genuine low-latency edge applications — not the radio.
URLLC as a service class. Reliability targets in the region of five nines, with latency budgets in single-digit milliseconds, using features like mini-slot scheduling and packet duplication. These carry a real capacity cost, which is exactly why they are a service class you request rather than a default everyone gets.
What this means when you are planning
Three practical implications come up in nearly every conversation we have on this.
First, check what your market actually has. SA availability varies enormously between operators in the same country, let alone between countries. A deployment plan that assumes slicing is available needs that confirmed operator by operator, in writing.
Second, latency claims need a measured baseline. The theoretical numbers in the standards describe the radio interface under specific conditions. The number your application experiences includes the core, the transport, the internet path and your own server. We have seen projects sized against a 1ms figure that measured 40ms end to end — and the radio was not the problem.
Third, private networks change the calculus. If your use case genuinely needs deterministic performance in a defined area — a factory, a port, a campus — a private 5G network gives you a standalone core you control, without waiting for a public operator's roadmap. That is increasingly where industrial 5G is actually being deployed.
The honest summary
5G non-standalone is a capacity and throughput upgrade. It is a real one, and for consumer mobile broadband it is most of the benefit.
5G standalone is an architectural change, and it is the version that supports the industrial and enterprise use cases the technology was sold on. If your plans depend on slicing, edge placement or guaranteed latency, you are planning for SA — and it is worth confirming that is what you will be running on before the business case is signed.
Our capacity building programme covers this ground in depth, from the 5G Primer through to protocol-level technical training, delivered in collaboration with the Qualcomm Wireless Academy.