Interactive Timeline
2G
legacyGSM / GPRS / EDGE - 1991
The network that gave the world SMS.
3G
legacyUMTS / HSPA / HSPA+ - 2001
The internet in your pocket - just about.
4G
deployedLTE / LTE-Advanced - 2009
All-IP. The generation that made smartphones indispensable.
5G
deployedNR (New Radio) - 2019
Not just faster - a different kind of network.
6G
researchIMT-2030 (Research Phase) - 2030
The network that merges the physical and digital worlds.
Circuit-switched voice core with a separate GPRS packet-switched overlay. The Base Station Subsystem (BSS) connects handsets to the core via BTS and BSC.
UTRAN (UMTS Terrestrial Radio Access Network) connects Node Bs via Radio Network Controllers to a split core: CS domain for voice, PS domain for data.
Flat, all-IP architecture. The EPC separates the control plane (MME) from the user plane (SGW/PGW). No circuit-switched domain - voice runs as VoLTE over the data plane.
Decomposed microservice architecture where every function exposes an HTTP/2 API. The control and user planes are fully separated (CUPS). The NRF enables service discovery across all functions.
6G architecture is not yet standardised. Research consensus describes a distributed, AI-native, cloud-native architecture with integrated sensing, non-terrestrial layers, and semantic communication.
What Changed in 2G
This is where it all began - the foundation for everything that followed.
What Changed in 3G
Spread-spectrum CDMA gives each user a unique code rather than a time slot, improving spectral efficiency and capacity.
3G introduced always-on IP connectivity as a first-class feature, not an overlay like GPRS.
Separate circuit-switched (voice) and packet-switched (data) cores ran in parallel.
Fast feedback loop between device and base station adapts modulation and coding rate per millisecond.
What Changed in 4G
OFDMA's subcarrier structure dramatically improves spectral efficiency and multi-user scheduling vs CDMA.
No more circuit-switched domain. Everything - including voice (VoLTE) - runs over IP.
Multiple antenna streams multiply throughput without consuming extra spectrum.
Radio resource management moved into the base station, flattening the RAN architecture and reducing latency.
Bonding multiple spectrum blocks to reach Gbps peak rates.
What Changed in 5G
EPC's monolithic nodes replaced by HTTP/2 microservices. Each function independently scalable and upgradeable.
Multiple virtual networks on one physical infrastructure - each with its own QoS profile, security, and topology.
64–256 antenna arrays form precision beams per user vs 4G's 4–8 antennas.
Hundreds of MHz of contiguous bandwidth per carrier - enabling 10+ Gbps peak in dense environments.
A new service tier designed from the ground up for 1ms latency and six-nines reliability.
UPF can be deployed at the network edge independently from the centralised core, enabling mobile edge computing.
What Changed in 6G
Opens entirely new frequency bands above mmWave, offering hundreds of GHz bandwidth per carrier for Tbps peak rates.
ML embedded into the waveform itself - not a layer on top. The radio interface learns and adapts continuously.
Communication signals double as radar - the network perceives and maps its physical environment in real time.
Tbps + <1ms enables multi-user holographic video - the convergence of physical and digital presence.
LEO satellites and HAPS become first-class access nodes in the same architecture as ground base stations.
A real-time virtual replica of the entire network enables AI-driven optimisation and predictive management at global scale.
Summary
2G replaced analogue 1G with digital transmission, which meant clearer calls, basic encryption, and - almost as an afterthought - SMS, tucked onto a signalling channel that was never meant to carry consumer traffic. It became one of the most lucrative accidents in telecoms history. Later extensions, GPRS and EDGE, bolted on packet data so phones could limp through WAP pages and early email, but speeds stayed in the kilobits: a single photo could take minutes to send. Decades on, 2G radios are still running quietly in the background of alarm systems, payment terminals, and fallback voice calls.
Rel-4 (EDGE) / GSM pre-3GPP - IMT - not formally classified
Summary
3G was the generation that made "mobile internet" mean something beyond a marketing slogan. Speeds jumped from kilobits to megabits almost overnight, and browsers, app stores, and video calls stopped feeling like party tricks and started feeling like utilities - though early adopters will remember the buffering wheel as a permanent fixture of daily life. HSPA and HSPA+ extensions kept stretching the ceiling, pushing theoretical speeds toward 42 Mbps by the end of the 3G era and buying the ecosystem time to build the smartphone habits that 4G would later cash in on. It's also the generation many carriers have now begun switching off, its spectrum handed over to 4G and 5G - a quiet irony given how much of the always-on app economy it first made possible.
Rel-99 (UMTS) through Rel-9 (HSPA+) - IMT-2000
Summary
4G LTE was the generation that finally tore up the old rulebook: instead of bolting data onto a voice network, it built a single all-IP network from scratch, with no circuit-switched domain left to maintain. OFDMA and MIMO antennas pushed peak speeds toward 1 Gbps in LTE-Advanced, but the bigger story was what that speed unlocked - Uber knowing exactly where you are, Instagram loading before you've finished the swipe, Netflix in your pocket, WhatsApp calls that don't count against your minutes. It's the generation the on-demand economy was actually built on, and more than a decade later it's still the workhorse most phones quietly fall back to the moment 5G coverage thins out.
Rel-8 (LTE) through Rel-13 (LTE-A Pro) - IMT-Advanced
Summary
5G's marketing promised flying cars and instant everything; the reality is more interesting, if less cinematic. Underneath the hype sits a genuinely new kind of network - a service-based core built from independently scalable microservices, three distinct service families for three very different jobs (fast phones, ultra-reliable machines, and huge swarms of low-power sensors), and network slicing that lets one physical network behave like several, each tuned for a different customer. Millimetre-wave spectrum delivers the multi-gigabit demos, though its short range means it mostly shows up in stadiums and dense city blocks rather than blanket coverage. Where 5G has actually landed, years into rollout, is quieter than the pitch: faster everyday browsing, viable cloud gaming, and the first mobile networks smart-city sensors and industrial IoT can genuinely depend on.
Rel-15 (5G NR Phase 1) through Rel-18 (5G-Advanced) - IMT-2020
Summary
6G doesn't exist yet - what follows is a set of research targets, not a shipping product, and history suggests the eventual reality will land somewhat short of the pitch (5G's own hype cycle is a useful reminder of that). ITU-R's IMT-2030 targets are expected to be finalised around 2027, with the first commercial networks pencilled in for 2030. The wish list is ambitious: sub-millisecond latency, terahertz spectrum offering vastly more bandwidth than mmWave, an air interface with machine learning built in rather than bolted on, and integrated sensing that turns the network itself into a kind of radar. Strip away the buzzwords and the underlying idea is simple - a network that doesn't just move your data, but has some awareness of the physical world it's moving through.
Rel-21+ (expected) - IMT-2030 (research targets)