Wi-Fi Protocols Compared: 802.11ax/be/ac (2026 Update)
Last updated: June 17, 2026.
A useful Wi-Fi protocols comparison starts by refusing the marketing headline. The box on the shelf shouts a single number — “up to 46 Gbps” — and that number is almost never the question you actually need answered. The honest question is narrower. Which standard gives your devices the airtime, the latency behaviour, and the band access they need, in the building you actually have, with the clients you actually own? That is the comparison this guide makes. We line up 802.11ac (Wi-Fi 5), 802.11ax (Wi-Fi 6 and 6E), and 802.11be (Wi-Fi 7). We explain what genuinely changed at the physical and MAC layers. And we stay ruthless about the gap between theoretical peak rates and what you will measure on a Tuesday afternoon.
This is a 2026 update because the landscape moved. Wi-Fi 7 hardware is now mainstream rather than early-adopter, and 6 GHz spectrum is widely available. Meanwhile the IEEE has begun work on 802.11bn — the basis of an eventual Wi-Fi 8 — explicitly aimed at reliability rather than another peak-rate record. We will treat each generation fairly, name the trade-offs out loud, and finish with a decision framework and checklist so you can map a standard to a scenario instead of chasing gigabits.
For industrial and IoT readers, there is a second thread running through this piece: when does newer Wi-Fi actually buy you determinism, and when should you reach for wired Time-Sensitive Networking instead? Wireless has closed part of that gap, but not all of it, and knowing the boundary is the difference between a plant network that works and one that intermittently does not. That boundary has shifted with Wi-Fi 7, but it has not disappeared, and a recurring theme below is being honest about exactly where it now sits.
How to actually compare Wi-Fi protocols
The peak PHY rate is the least informative number on the spec sheet, yet it dominates every comparison chart. To compare Wi-Fi protocols properly you have to break that headline rate into its ingredients, because each ingredient behaves differently in the real world.
A link’s raw rate is roughly the product of four things. They are channel width, modulation density (the MCS), the number of spatial streams, and the MAC efficiency that decides how much of the air is wasted on overhead. Multiply the first three and you get the physical-layer maximum the marketing team prints. Apply the fourth, plus interference and distance, and you get throughput a human would recognise.
Consider each lever in turn. Channel width is the most direct: doubling from 80 MHz to 160 MHz roughly doubles the rate, and Wi-Fi 7’s 320 MHz doubles it again. But you only get that width where spectrum is clean enough to use a channel that wide, which in practice means the 6 GHz band. Modulation packs more bits into each symbol: 802.11ac tops out at 256-QAM, 802.11ax reaches 1024-QAM, and 802.11be reaches 4096-QAM. Each step up demands a cleaner signal, so the densest modulation only appears close to the access point with little interference. Spatial streams use multiple antennas to send parallel data streams. More streams need more radio chains in both the client and the AP. Most phones and IoT devices ship with only one or two streams, regardless of what the router supports.
That last point is the quiet killer of peak-rate dreams. A router advertising eight streams and 4096-QAM still talks to a two-stream phone at, at best, two-stream rates. The link is negotiated to the lesser device. Your slowest, oldest client sets the ceiling for its own traffic, and a crowded channel sets it for everyone.
Then there is MAC efficiency — how cleverly the standard shares the air among many devices. This is where the modern generations earn their keep, and where a fair Wi-Fi protocols comparison should spend most of its attention. Wi-Fi 5 essentially serves one device at a time per channel and waits its turn. Wi-Fi 6 introduced OFDMA, which subdivides a channel so several clients transmit at once. Wi-Fi 7 added Multi-Link Operation, letting one device use several bands simultaneously. None of these touch the headline rate, yet all of them transform behaviour in a busy room — which is exactly where most networks actually live.
So the comparison framework is: width × modulation × streams gives the spec-sheet ceiling; MAC efficiency, interference, distance, and client capability decide what you keep. Hold that mental model and every generation below slots into place.
It helps to make the multiplication concrete. Imagine a router that advertises eight spatial streams, 4096-QAM, and a 320 MHz channel — the full Wi-Fi 7 stack. The peak number on the box assumes all three maxed at once. But your phone has two streams, so its share of the link is one quarter of the stream dimension. It rarely sustains 4096-QAM unless it is sitting on the desk beside the AP, so the modulation dimension drops too. And a 320 MHz channel only exists in 6 GHz, where the phone’s range is shorter, so distance erodes both modulation and reliability. Three multipliers, each quietly reduced, and the realistic figure lands far below the headline. None of this is a defect — it is simply how a shared, negotiated medium works — but it is why the spec-sheet number should never anchor a capacity plan.
There is one more dimension the headline rate ignores entirely: how many devices want the air at the same time. A single client on an empty channel is the easy case every benchmark uses. A real network has dozens of devices competing, and the standard’s scheme for arbitrating that competition — one-at-a-time, OFDMA, or multi-link — often matters more to perceived performance than any raw-rate figure. That is precisely why the MAC-layer comparison below carries more practical weight than the PHY-rate column.
The standards explained
The figure above is the family tree. Read it left to right and notice the pattern: each generation adds spectrum, modulation density, and — increasingly — MAC-layer intelligence, while the very newest work pivots toward reliability rather than raw speed.
802.11ac (Wi-Fi 5)
802.11ac, ratified in 2014 and branded Wi-Fi 5, is still the floor of most installed networks. It operates only in the 5 GHz band, supports channel widths up to 160 MHz, and reaches 256-QAM modulation. Its signature MAC feature was downlink multi-user MIMO, which let the access point transmit to several clients at once on the downlink — a real improvement over Wi-Fi 4, but a one-way one.
Wi-Fi 5’s limits are instructive precisely because they motivated everything after it. It has no access to the congested 2.4 GHz band’s relief valve, no uplink multi-user capability, and no mechanism to subdivide a channel among small transmissions. In a dense environment — an office, an apartment block, a factory floor with dozens of sensors — Wi-Fi 5 spends much of its time waiting for the channel to clear. For a handful of high-throughput clients with line of sight to the AP, it remains perfectly serviceable, and a great deal of working hardware still runs on it. For density or low-power IoT, it shows its age.
There is a useful lesson buried in Wi-Fi 5’s design. It chased throughput — wider channels, more streams, downlink multi-user — and largely succeeded for the single-client case. What it could not do was make a crowded network feel good, because every device still fundamentally took turns. That gap is exactly what motivated Wi-Fi 6 to pivot from “make one link faster” to “make many links coexist.” If you understand why Wi-Fi 5 struggles in density despite its respectable peak rate, you understand the entire trajectory of the standards that followed. The story of modern Wi-Fi is the story of contention, not throughput.
802.11ax (Wi-Fi 6/6E)
802.11ax, branded Wi-Fi 6, is the generation where the focus shifted from peak speed to efficiency under load, and it is arguably the most consequential upgrade in the family. It operates in 2.4 GHz and 5 GHz, and the 6E designation extends it into the new 6 GHz band where regulators permit. It pushes modulation to 1024-QAM for a modest rate bump close to the AP.
The substance is in the MAC layer. OFDMA subdivides each channel into resource units, so the AP can serve many small clients in a single transmission instead of one at a time. That is transformative for environments full of IoT sensors and phones sending tiny packets. Uplink and downlink multi-user MIMO make the parallelism bidirectional. BSS coloring tags transmissions so a device can distinguish its own network from a neighbour’s on the same channel and transmit anyway, rather than backing off needlessly. And Target Wake Time (TWT) lets a device negotiate scheduled wake windows, sleeping the rest of the time — a genuine battery-life win for sensors, and the first standard feature with a determinism flavour.
For the industrial IoT communication stack, Wi-Fi 6 is often the pragmatic sweet spot. It handles client density, it lets battery devices sleep predictably, and 6E opens a clean band away from legacy interference. Much of what people credit to Wi-Fi 7 actually arrived with Wi-Fi 6.
802.11be (Wi-Fi 7)
802.11be, branded Wi-Fi 7 and finalised in 2024, is the current peak of the family. It keeps the three bands (2.4, 5, and 6 GHz), doubles channel width to 320 MHz, lifts modulation to 4096-QAM, and scales up to 16 spatial streams in principle. Those three levers together are why the headline numbers leapt into the tens of gigabits.
But Wi-Fi 7’s defining features are architectural, not just bigger. Multi-Link Operation (MLO) lets a single client use multiple bands simultaneously — aggregating them for throughput, or duplicating traffic across them for reliability and lower latency. Preamble puncturing lets the radio use a wide channel even when part of that spectrum is occupied by interference, by “punching a hole” around the busy sub-channel instead of abandoning the whole width. And the standard’s later Release 2 work adds explicit low-latency mechanisms aimed at real-time traffic. We will unpack MLO properly in the latency section, because it is the feature that matters most for the comparison.
Preamble puncturing deserves a moment on its own, because it changes the economics of wide channels in messy spectrum. Older standards treated a wide channel as all-or-nothing: if any sub-channel was occupied, the radio either fell back to a narrower width or waited. Puncturing lets the transmitter mask out just the busy slice and use the rest of the wide channel anyway. On a real industrial site, where a stray interferer often parks on one 20 MHz slice, this is the difference between keeping a usable 240 MHz and collapsing back to 80 MHz. It is a quiet feature with an outsized effect on sustained throughput in the real world.
The honest framing: Wi-Fi 7’s peak rate is mostly a 6 GHz, 320 MHz, short-range story. Its reliability and latency improvements are the part that travels — and the part that makes Wi-Fi 7 industrial 2026 deployments worth a serious look.
On the horizon: Wi-Fi 8 / 802.11bn
The IEEE 802.11bn task group is developing Ultra High Reliability (UHR), the basis of an eventual Wi-Fi 8. Crucially, it is not chasing a new peak-rate headline. The stated goals centre on reliability and worst-case behaviour: lower tail latency, more consistent throughput at the cell edge, and coordination between access points so neighbouring APs cooperate rather than collide. For anyone who cares about deterministic wireless, this is the more interesting direction — speed has outrun most applications, while predictability has not. Wi-Fi 8 is years from products, but its priorities tell you where the technology’s centre of gravity is moving.
The multi-AP coordination piece is the part to watch. Today, neighbouring access points on the same channel mostly treat each other as interference and back off, wasting airtime. UHR aims to let them schedule and even transmit cooperatively, so a dense deployment behaves more like one coordinated system and less like a crowd of competitors. If that lands, it would do for between-AP contention what OFDMA did for within-cell contention — and it is precisely the multi-AP behaviour that large industrial and campus deployments struggle with most. For planning purposes today, the takeaway is simple. Do not wait for Wi-Fi 8 to solve a problem Wi-Fi 6 or 7 already addresses. But do expect the reliability story, not the speed story, to define the next decade of Wi-Fi.
Head-to-head comparison
The table below is the side-by-side. Treat the peak PHY rate column as a theoretical ceiling under ideal conditions — widest channel, densest modulation, maximum streams, no interference. Real-world throughput is routinely a third to half of these figures, and far lower with distance, walls, or a busy channel.
| Attribute | 802.11ac (Wi-Fi 5) | 802.11ax (Wi-Fi 6/6E) | 802.11be (Wi-Fi 7) |
|---|---|---|---|
| Bands | 5 GHz | 2.4 / 5 / 6 GHz | 2.4 / 5 / 6 GHz |
| Max channel width | 160 MHz | 160 MHz | 320 MHz |
| Modulation (max) | 256-QAM | 1024-QAM | 4096-QAM |
| Peak PHY rate (spec) | ~3.5 Gbps | ~9.6 Gbps | ~46 Gbps |
| Real-world expectation | A fraction of peak; one or two streams typical | Higher efficiency under load; OFDMA helps density | High peak only on clean 6 GHz wide channels |
| Key MAC features | DL MU-MIMO | OFDMA, UL/DL MU-MIMO, BSS coloring, TWT | MLO, preamble puncturing, low-latency R2 |
| Latency posture | Best-effort | Better under load; TWT scheduling | Lowest tail latency via MLO + R2 |
Read the matrix as a progression of capability surface, not just speed. Wi-Fi 5 is a single-band, best-effort link. Wi-Fi 6 adds bands and airtime efficiency. Wi-Fi 7 adds simultaneous multi-band operation and a latency story. The peak-rate jump is real but conditional; the MAC-feature jump is what you feel every day.
This is also the cleanest way to frame 802.11ax vs 802.11be and the broader Wi-Fi 6 vs Wi-Fi 7 question. The gap between them is less about a bigger number and more about MLO, wider channels, and Release 2’s latency work. If your environment cannot use 320 MHz channels — and most cannot, outside clean 6 GHz — then Wi-Fi 7’s advantage over a good Wi-Fi 6E deployment narrows to MLO and latency rather than throughput.
What actually changes latency and reliability
Throughput sells routers; latency and reliability run factories and video calls. Four features carry most of the weight here, and understanding them is the heart of any serious Wi-Fi standards explained discussion.
OFDMA (Wi-Fi 6 onward) is the first. By splitting a channel into resource units, it lets many devices transmit in the same time slot instead of queuing. The effect is not a faster single link — it is dramatically lower latency under contention. A room with fifty IoT sensors firing small packets behaves completely differently on OFDMA than on Wi-Fi 5’s one-at-a-time scheme. For dense IoT, this is the single most valuable latency improvement in the family’s history. The reason is mechanical: under the old scheme, a tiny packet still seized the whole channel for its turn, so fifty small senders meant fifty serial turns and fifty stacking delays. OFDMA collapses many of those turns into one, so the delay each device experiences stops growing linearly with the crowd. That is why dense networks, not headline benchmarks, show OFDMA’s value most clearly.
Target Wake Time (TWT) is the second, and it is subtler. TWT lets a device and the AP agree on scheduled wake windows. For battery sensors this saves power, but the determinism angle matters more here: scheduled access means a device’s transmissions are planned rather than contending randomly, which reduces jitter for traffic that fits the schedule. It is not full time-division determinism, but it is a meaningful step toward predictable airtime.
BSS coloring (Wi-Fi 6 onward) attacks a different problem: in dense deployments, devices waste time deferring to transmissions from neighbouring networks on the same channel. Coloring tags each network so a device can recognise an “other-network” signal and transmit anyway under the right conditions, recovering airtime that older standards threw away. More usable airtime means lower queueing delay.
These three features compound rather than compete. A dense Wi-Fi 6 deployment uses OFDMA to pack many small flows into shared airtime, TWT to schedule predictable wakeups for sensors, and BSS coloring to stop wasting time deferring to neighbours. Each one attacks a different source of delay — contention, random wakeups, and over-cautious backoff respectively. Together they explain why a well-configured Wi-Fi 6 network can feel dramatically more responsive than a Wi-Fi 5 one even when the raw link rate is similar. The latency win is structural, not a faster pipe. That is the point worth internalising: most of the everyday improvement in modern Wi-Fi comes from spending airtime more wisely, not from moving bits faster.
Multi-Link Operation (MLO, Wi-Fi 7) is the marquee reliability feature, and it deserves the diagram.
A Wi-Fi 7 client can associate over several bands at once rather than picking one. The MLO manager then chooses a strategy per traffic flow. For bulk transfer it aggregates links to add their capacity together. For latency-sensitive or critical traffic it can duplicate a packet across two links, so if one band suffers a burst of interference, the copy on the other arrives on time. That duplication is the key reliability mechanism: it converts the worst-case latency of a single noisy channel into the better of two channels, which is exactly the behaviour real-time control needs.
MLO is why Wi-Fi 7 industrial 2026 is a genuine conversation rather than a marketing line. Combined with Release 2’s explicit low-latency scheduling, it brings wireless meaningfully closer to bounded latency — closer than any prior Wi-Fi generation. It does not make Wi-Fi deterministic in the hard, safety-rated sense, but it shrinks the tail of the latency distribution, and for a large class of monitoring and soft-real-time applications that is enough.
A worked intuition makes the difference clear. On a single channel, latency is dominated by its worst moments — the burst of interference, the retransmission, the contention spike — and those rare bad moments are exactly what a deadline-driven application fears. With MLO duplication across two independent bands, a packet only misses its deadline if both links happen to be bad at the same instant, which is far less likely than either being bad alone. Statistically, you are taking the better of two draws on every critical packet. The average latency may barely move; the 99th-percentile latency — the number that actually breaks real-time systems — improves substantially. That shift in the tail, not the average, is the real industrial value of Wi-Fi 7.
It is worth being precise about what MLO does not do. It does not provide a formal worst-case guarantee. It does not coordinate between separate access points; that is Wi-Fi 8’s territory. And its benefit depends on the two links being genuinely independent, since duplicating across two congested bands helps little. Used well, on a clean 6 GHz link paired with a 5 GHz link, it is a powerful reliability tool. Sold as “deterministic Wi-Fi,” it is overstated.
Industrial and IoT considerations
Industrial requirements differ from home and office requirements in one decisive way: they care about the worst case, not the average. A consumer cares that the video usually streams; a robot cell cares that a control message never arrives 50 ms late. That difference reshapes the whole comparison.
For genuinely hard real-time, safety-rated control — motion control, interlocks, anything where a missed deadline is a hazard — the answer in 2026 is still wired Time-Sensitive Networking or a fieldbus, not Wi-Fi. TSN over Ethernet provides bounded, deterministic delivery with formal guarantees that no current Wi-Fi standard matches. If you are designing that layer, the TSN reference architecture is the right starting point, and wireless should stay in the soft-real-time and monitoring tiers above it.
A concrete example sharpens the boundary. Mobile robots — AGVs and AMRs roaming a warehouse — are a classic wireless case, because cabling a moving vehicle is impossible. Their control loops are usually soft-real-time: a late navigation update slows the robot or triggers a safe stop, but it is not an instant hazard the way a motion-control interlock is. This is exactly the tier where Wi-Fi 7’s MLO earns its keep. Duplicating navigation traffic across 5 GHz and 6 GHz keeps the robot responsive through the dead spots and interference bursts that a single link would stumble on. The safety-rated emergency-stop function, by contrast, still belongs on a dedicated safety system, not on best-effort Wi-Fi. Splitting traffic by criticality — soft-real-time on resilient Wi-Fi, hard safety on wired or dedicated channels — is the pattern that makes wireless robotics dependable.
For soft-real-time and high-density IoT — telemetry, condition monitoring, AGV coordination with reasonable tolerances, AR maintenance guidance — modern Wi-Fi is increasingly viable. Wi-Fi 6’s OFDMA and TWT handle sensor density and predictable wakeups. Wi-Fi 7’s MLO duplication trims the latency tail enough for many soft deadlines. The 6 GHz band is a quiet hero here: it is wide, comparatively uncongested, and free of legacy 802.11b/g/n clutter, which makes it the cleanest place to run latency-sensitive wireless on a busy site.
Regulation shapes what you can actually deploy. The 6 GHz band that makes Wi-Fi 6E and Wi-Fi 7 compelling is not opened uniformly worldwide. Different regions allow different portions of it, at different power levels, and some require automated frequency coordination for outdoor or higher-power use. A design that assumes full 320 MHz availability may be legal in one country and not another. For multi-site industrial rollouts this matters: verify the local spectrum rules before standardising on a band plan, or you will ship a configuration that cannot be certified at half your sites. Treat 6 GHz availability as a per-region question, not a global given.
Coexistence is the underrated risk on a factory floor. The 2.4 GHz band shares space with Bluetooth, Zigbee, microwave ovens, and legacy equipment; the 5 GHz band shares with
