The 6G physical layer is starting to produce hard numbers. The bigger commercial question — how operators actually move from 5G to 6G — remains unresolved.
Madrid delivered meaningful progress on the radio interface, but little closure on the migration architecture. For chipset designers, the direction is becoming clearer. For network planners, the most consequential decision has moved to December.
That contrast is becoming increasingly important as 3GPP works toward the Release 21 milestones and its contribution to the ITU-R IMT-2030 process.
1. The radio is becoming more concrete
Three of the four physical-layer items RAN#112 had set as Q3 homework came back with answers at RAN1#126 in August and were reported at the Madrid meeting.
Uplink BG3 is in place, with a carefully chosen floor.
Support is mandatory with capability signaling, except for UEs whose maximum uplink channel bandwidth is ≤20 MHz in any supported band. The parameters are largely set: 44 information columns, a maximum lifting size of 192, a mother rate around 2/3 and a maximum code block size of 8,448 bits — matching BG1 — under a constraint that the lifted parity-check matrix’s total number of ones stays at or below BG1 at rate ≥1/2. Structurally, submatrix E is lower-triangular rather than identity; the threshold I remains open.
The 20 MHz line matters because it limits mandatory BG3 support for devices with low maximum uplink bandwidth, reducing the likelihood that the new decoder requirement becomes a universal silicon burden. In other words, 6GR is adding capability while trying to keep the low end of the device ecosystem from paying the full complexity cost.
SSB continues 5G’s frame structure rather than breaking it.
RAN1 agreed on 20 RB, identical to the NR SSB, with PSS at 12 RB. The time-domain length is a working assumption of 4–7 symbols, compared with 5G’s fixed four, while default initial-search periodicity is being evaluated across 20/40/80/160 ms, with the longer options premised on repetition and UE-side combining.
This is less about SSB itself than about migration economics.
Keeping the SSB grid aligned with NR makes shared-carrier operation more tractable and is consistent with RAN1’s separate agreement to align NR and 6GR resource-block boundaries on an MRSS carrier.
The variable time-domain length also gives RAN1 another lever for the coverage trade-off that becomes increasingly important at higher frequencies, including around 7 GHz: longer SSB transmission can improve link budget, while longer periodicity can reduce network and device energy consumption. That trade-off has not yet been fully settled.
Higher-order modulation was scoped down, not expanded.
1024QAM is supported in the uplink for FWA use, with RAN1 recommending it. There is no consensus on 4096QAM support in Release 21. Qualcomm resisted the original “not supported” wording, leaving the door formally open, while MediaTek argued that 4096QAM does not outperform 1024QAM and Huawei did not support it.
Constellation shaping was terminated.
June had instructed RAN1 to decide or stop. Support was split roughly 9–18 companies for and 7–15 against, depending on the scheme, and the work ended without agreement, with misaligned simulation assumptions cited as a factor.
The outcome reinforces a broader 6GR design direction: avoiding multiple competing options for the same function where the performance gain is not sufficiently clear.
That matters for device roadmaps because every additional optional capability can translate into modem complexity, testing and interoperability requirements.
Taken together, these decisions show a physical layer that is becoming more disciplined: add capability where there is a clear use case, constrain the hardware burden, and avoid proliferating alternatives without a sufficiently compelling performance gain.
2. MRSS: coexistence is becoming measurable
The next question is how this new radio actually lives alongside 5G.
The plenary endorsed efficient sharing of at least NR and 6GR downlink control-channel resources in 6GR, targeting the downlink control overhead RAN1 had tabled for shared-resource cases.
The rationale is concrete: separate CORESETs for NR and 6GR in a 10 MHz FDD carrier can add more than 14% overhead, rising to 23% in the worst case.
That creates a measurable cost for running separate control structures.
Operators asked for dynamic sharing and received an efficient approach, deliberately leaving the mechanism flexible for RAN1 to determine. RAN1 did not adopt additional operator proposals covering shared reference signals and multiplexed uplink control and sounding.
On the mechanism side, RAN1 has already agreed that 5G and 6G CORESETs may overlap, providing a route to reducing the control overhead.
So MRSS has moved from an agreed baseline to an agreed baseline with a measurable target.
That is meaningful progress, but it is still only a partial answer to the operator problem. The target addresses downlink control overhead; it does not yet define how operators will manage mobility, uplink control, reference signals and the broader operational complexity of running NR and 6GR together.
The significance is therefore bigger than the percentage itself: 6GR is being designed with coexistence in mind rather than treating 5G as something that simply disappears when 6G arrives.
3. But migration architecture remains unresolved
This is where the Madrid meeting produced the biggest unanswered question. Migration down-selection failed again.
The Chair opened by stating that the discussion should focus on down-selection, and the operator tally was completed.
The result was 18 operators supporting Option 1 — 6G-anchored dual connectivity — eight supporting MRSS only, six supporting Option 3 and three supporting Option 2, with some operators supporting more than one option.
No wording survived objection. The decision therefore moves to RAN #114 in Boston in December, which will also address split bearer for Options 1 and 2.
A simplified Option 3 remains on the table: no expected RAN impact, two single registrations, network-controlled mobility only, voice-over-NR continuity with 6G data as the single use case, with SA ultimately making the final call.
The unresolved migration question matters because the options are not minor implementation variations. They imply different relationships between 5G and 6G during the transition.
And this is where the architecture problem becomes apparent.
The migration decision and the 6G core architecture are closely connected. Vodafone made its support for Option 1 conditional on a lightly evolved 5G core, while the SA WG2 architecture work is progressing toward the same timeframe.
Meanwhile, Deutsche Telekom, T-Mobile US, Telstra and Jio are among the operators supporting an MRSS-only direction. That means the debate is also about how much of the existing 5G architecture operators should carry forward, how much new 6G functionality they should introduce, and where the boundary between evolution and a new architecture should sit.
The radio is converging faster than the migration architecture.
That is the central outcome from Madrid. Continued disagreement leaves operators without a neutral architectural choice; it increases the practical importance of the MRSS-only path while the other options remain under debate.
December is therefore becoming the critical decision point. The March 2027 Release 21 Stage-1 freeze and the subsequent specification milestones leave limited room for a prolonged architecture debate before the detailed work has to move forward.
4. The external clock is already running
One major question, however, was settled unambiguously. 3GPP has agreed that its IMT-2030 candidate will be based on 6G Radio, rather than bundling NR into the submission as LTE was bundled for IMT-2020.
The technology will be submitted as “6G”, with the footnote “Developed by 3GPP as 6G, Release 21 and beyond.”
The associated milestones put the work on a tight schedule:
- February 2027: high-level presentation at the ITU-R WP 5D workshop
- June 2027: initial submission template
- June 2028: partial self-evaluation
- February 2029: final self-evaluation
The February 2029 evaluation milestone comes only one month before the March 2029 ASN.1 freeze. That leaves very little schedule margin between 3GPP’s final specification milestones and the ITU-R evaluation timeline. That’s why the unresolved migration question matters beyond the RAN1 room.
The radio is converging. The migration path isn’t.
Madrid shows two very different speeds in 6G standardization. At the radio level, the picture is becoming increasingly concrete: coding, SSB structure, modulation and shared-resource efficiency are moving from research questions toward engineering parameters. The direction is also increasingly compatible with coexistence with NR.
At the network level, the harder question remains open: how operators move from today’s 5G architecture to a 6G network without creating unnecessary duplication, complexity, or stranded investment.
That distinction matters. A technically mature 6G air interface does not by itself define a commercially deployable 6G network. The next critical step is the migration decision — and how that decision aligns with the 6G core architecture before the Release 21 timeline removes further room for maneuver.