IMT-2030 Capabilities: Nine Enhanced, Six New

You have been through this chain once already. M.2083 set the IMT-2020 framework in 2015, M.2410 turned it into minimum technical performance requirements, M.2412 defined the test environments those requirements were evaluated in, and M.2150 recorded what was accepted at the end. Everything argued about in between — the 20 Gbit/s, the 1 ms, the million devices — came from the middle two documents, not the first.

ITU-R M.2160, approved at RA-23 in November 2023, is the M.2083 of this cycle. It defines fifteen capabilities: nine carried over from IMT-2020 with raised targets, six with no predecessor at all.

The split is the whole story. Raise the nine and you have 5G with more spectrum. The six are the part that changes what the system is for.

---

The full list

Two structural points before the detail.

Nine of the fifteen carry a number — the eight quantitative enhanced ones plus positioning. The remaining six are named without targets because nobody has agreed a measurement that survives being turned into a pass or fail.

None of these are requirements. M.2160 says "could be" and "possible examples", exactly as M.2083 did. You already know what happened to M.2083's numbers when M.2410 bound them: 20 Gbit/s downlink became an analytical calculation at maximum bandwidth and eight layers, and 1 ms became one-way user plane latency in the URLLC test environment only. Expect the same treatment here.

---

Chapter 1 — The nine enhanced capabilities

The rate capabilities

Peak rate 20 → 50–200 Gbit/s. User experienced rate 100 → 300–500 Mbit/s. Spectrum efficiency 1.5–3× IMT-2020.

Read those three together and the shape is familiar. Peak rate tracks aggregated bandwidth and layer count, so it scales with whatever spectrum gets allocated — it says more about WRC outcomes than about the radio. Note also that IMT-2020's headline 20 Gbit/s was downlink, against 10 Gbit/s uplink, and whether IMT-2030 keeps that asymmetry is a requirements-report question, not a framework one.

Spectrum efficiency is the honest number, and it is 1.5–3× over a full generation. You know why: link adaptation, coding and MIMO processing are already close enough to the bound that the remaining gains come from spatial reuse and from operating points, not from a better waveform. Anyone briefing you on a tenfold 6G capacity gain is quietly counting bandwidth.

Capacity and density

Area traffic capacity 10 → 30–50 Mbit/s/m². It falls out of spectral efficiency, bandwidth and site density, so it is closer to a deployment statement than a radio one.

Connection density 10⁶ → 10⁶–10⁸ per km². The widest bracket in the table, and worth reading as a hedge rather than a target. Recall how M.2410 bound the 5G figure — one million devices per km² at a defined packet arrival rate and payload in the mMTC environment. A hundredfold on top of that binding is a different problem from a hundredfold in the abstract, and which binding the requirements report chooses will decide whether the top of this range is aspiration or engineering.

Latency and reliability

Latency 1 ms → 0.1–1 ms. The top of the range did not move, which is the interesting half. M.2160 is not promising sub-millisecond generally; it is opening the floor. The microsecond figures in circulation are from neither document.

Reliability 1−10⁻⁵ → as far as 1−10⁻⁷. M.2410 defined the 5G figure tightly — success probability for a 32-byte layer-2 PDU inside 1 ms at coverage-edge quality — and 10⁻⁷ inherits that framing rather than replacing it. Two more orders of magnitude are bought with redundancy, repetition or a lower operating MCS, all of which you pay for in spectral efficiency. Reliability and capacity trade; they always have.

Mobility and security

Mobility 500 → 1000 km/h. Aviation and high-speed rail. The binding constraint is Doppler and the resulting phase noise and ICI budget, not handover rate — which is why this interacts with numerology choice more than with mobility procedures.

Security and resilience is the ninth enhanced capability and carries no value. It is also one of the four overarching aspects, which is the first sign of the list-overlap problem covered below. Resilience here means continuing through a disruptive event and recovering quickly — a system property, not a radio measurement, and unquantifiable for that reason.

---

Chapter 2 — The six new capabilities

None of these have an IMT-2020 counterpart. They were absent, not weaker.

Coverage

The one most often misfiled as enhanced, since coverage has always mattered operationally. It has — but never as a formal capability. M.2083's set has no coverage entry, and every capability in it describes served conditions rather than the edge.

M.2160 defines it as cell-edge service range from link-budget analysis. Putting a link budget in a capability table is a corrective to three generations of peak-condition metrics, and it is the capability most directly threatened by the upper mid-band, where the geometry no longer closes the budget on its own.

Positioning

1–10 cm, and the only new capability with a number. 5G positioning requirements were specified in metres, with sub-metre only under Release 16/17 enhancements and favourable geometry.

Centimetre accuracy needs bandwidth, geometry and inter-site synchronisation simultaneously. The physics permits it; the deployment tolerances are the problem. Expect the widest best-case-to-median gap of anything in the table.

Sensing-related capabilities

Range, velocity and angle estimation, detection, localisation, imaging and mapping, provided through the radio interface. This is the capability under ISAC.

No target value, and it is not obvious what a single one would mean — sensing performance is a joint function of bandwidth, geometry and target cross-section, so a scalar without those three stated is unusable. Watch whether the requirements report defines a bounded case or leaves it qualitative.

Applicable AI-related capabilities

Distributed data processing, distributed learning, AI computing, model execution and inference, as functions the network provides.

The framing matters: not the network using models internally, which 5G-Advanced already does through Release 18, but the network offering compute and inference alongside connectivity. That is a claim about what a mobile network is, and it is why AI-native describes a design stance rather than a feature set.

Sustainability

Network and device emissions and environmental impact across the full lifecycle.

Two things make it stronger than the energy-efficiency language you already know from M.2410, where efficiency was a qualitative capability tied to sleep ratio and slot duration. It includes devices, and it includes manufacture and disposal rather than only operating power. Also unquantified — the system boundary would have to be agreed first.

Interoperability

A radio interface built on member-inclusivity and transparency so that functions work between different entities of the system.

The least technical entry and the most political. It is in a capability set because the Open RAN decade established that multi-vendor operation is either designed in or litigated afterwards. Its presence here is ITU-R putting that argument inside the specification phase.

---

Where this actually bites: the requirements report

M.2160 was never testable, and neither was M.2083. The document that decides whether a candidate RIT qualifies is the minimum technical performance requirements report, and it has now landed.

ITU-R WP 5D finalised a draft in February 2026 with 20 technical performance requirements, seven of them new to 6G, spanning peak rate and peak spectral efficiency, user experienced rate, latency, reliability, connection density, mobility, positioning accuracy, sensing, energy efficiency and coverage. Approval is expected when the parent study group meets in December.

The values are not public — TIES access until approval. So the honest position today is: shape known, thresholds not.

That distinction is the practically important one, because you have seen what the binding does. A capability becomes a requirement, the requirement becomes a test environment in the evaluation guidelines, and the test environment is what vendors self-evaluate against and therefore what gets optimised. The list decides what gets simulated, and what gets simulated decides what ships. Which of these six new capabilities acquires a bounded definition in that report is the single thing worth tracking this year.

Three lists, not one

M.2160 keeps three lists apart and most summaries merge them — including an earlier version of our own 6G overview.

Six usage scenarios — immersive communication, massive communication, hyper reliable and low-latency communication, ubiquitous connectivity, AI and communication, integrated sensing and communication. The successors to eMBB, mMTC and URLLC, with three additions.

Four overarching aspects — sustainability, security and resilience, connecting the unconnected, ubiquitous intelligence. Design principles across all scenarios; M.2083 had no equivalent.

Fifteen capabilities — what gets measured, and the only list the requirements report acts on.

Sensing, security and sustainability each appear on two lists with different meanings on each. The overlap is deliberate in the source and becomes noise the moment the three are flattened into one bullet list of features.

Reading the table without being misled

"Could be" is not "shall be". Same status as M.2083's figures before M.2410 bound them.

Ranges are ranges. Quoting 200 Gbit/s or 10⁸ per km² as the 6G number is taking the top of a bracket.

Nothing meets all fifteen simultaneously. They trade — reliability against efficiency, latency against scheduling flexibility, coverage against rate. The table describes the technology family's envelope, not one cell's operating point.

No number is not the same as no substance. Six are unquantified because the measurement is genuinely unsolved. A premature metric would be worse than none, and the ones that acquire numbers are where real progress is visible.

---

The mental model

M.2160 is this cycle's M.2083. Framework, not requirements. The binding happens downstream, and you have watched it happen once.

Fifteen capabilities: nine enhanced, six new. Only nine carry numbers — the eight quantitative enhanced ones plus positioning at 1–10 cm.

Coverage is new; security and resilience is enhanced. The pair most often swapped in both directions.

Spectrum efficiency 1.5–3× is the honest figure. Peak rate scales with allocated bandwidth, not with radio design.

Twenty requirements, seven new, values under TIES. Shape known, thresholds not, until December.

What gets bound gets simulated, and what gets simulated ships. That is why the list is worth reading at source.

Further reading

• Recommendation ITU-R M.2160-0 — the framework, and the source of every figure above
• ITU-R WP 5D IMT-2030 programme page — capability definitions and current status
• IMT-2030: technical requirements for the 6G future — ITU on the twenty minimum requirements
• Recommendations ITU-R M.2083, M.2410 and M.2412 — the IMT-2020 framework, its minimum requirements and its evaluation guidelines, for the comparison column and for what binding looks like
• What is 6G? for the timeline, and what will actually change for the engineering consequences