Chapter One

What is 6G Core? (The Simple Version)

The core network is the brain of every mobile network. When you make a call, send a message, or watch a video, the core network is the invisible system that authenticates you, finds the right path for your data, applies your subscription rules, and connects you to the internet or another user. In 5G, this brain was rebuilt from scratch as the 5G Core (5GC), defined in TS 23.501. Now, 3GPP is designing the next brain: the 6G Core (6G CN), studied in TR 23.801.

Think of it like this

The 5G Core is like a modern hospital — each department (network function) has a specialty, they communicate through a shared computer system (SBA), and you get treated based on your insurance plan (subscription). The 6G Core is like upgrading that hospital with: an AI that predicts which patients will arrive tomorrow, a “digital twin” simulation to test new treatments, a built-in radar that monitors everyone’s health in the building, and a computing lab that can run any medical analysis on-site instead of sending samples away.

24
Key Issues in TR 23.801
8
Work Tasks
V0.4.0
Latest Version
Rel-20
3GPP Release

But here’s what makes this article different: we’re not guessing about 6G. We’re reading the actual 3GPP study — TR 23.801-01 V0.4.0, published February 2026 — the official document where 3GPP SA2 (the system architecture working group) is designing the 6G Core. This document has 24 key issues, 8 work tasks, and dozens of proposed solutions. We’ll decode all of them.

• • •
Chapter Two

5G Core vs 6G Core: The Big Picture

Before diving into the 24 key issues, let’s understand the fundamental architectural shift. The good news: 6G Core does not throw away 5G Core. TR 23.801 explicitly states:

“The framework of SBA specified for 5GC is assumed as a starting point for discussion.”

— TR 23.801 V0.4.0, Section 4.1 (Architectural Assumptions)

So 6G Core evolves from 5G Core. But the evolution is profound. Here’s the side-by-side:

5G CORE (TS 23.501) Service-Based Architecture AMF SMF AUSF UDM PCF NRF NSSF NEF SBA BUS (HTTP/2 REST APIs) NWDAF (analytics add-on) MEC (separate edge platform) UPF CHARACTERISTICS: Architecture: Service-Based (SBA) AI/ML: NWDAF (Rel-16 add-on, observes) Sensing: Not supported Computing: MEC separate platform NTN/Satellite: Bolt-on (Rel-17+) Slicing: NSSAI-based, static Intent-based: Not supported Spec: TS 23.501 / TS 23.502 6G CORE (TR 23.801) SBA + AI-Native + Intent-Based INTENT LAYER (KI#18) — AI Agents 6G-AMF 6G-SMF 6G-PCF 6G-NRF/NEF Sensing KI#20 Compute KI#22 AI/ML KI#18/19 Data Fwk KI#21 ENHANCED SBA (KI#2) + MODULAR NAS (KI#1) 6G-UPF + Computing Site User Plane + Edge Compute (KI#4/22) Native NTN/Satellite (KI#23) IoT Enablers (KI#24) CHARACTERISTICS: Architecture: SBA (evolved) + Intent Layer AI/ML: Native AI agents (KI#18), AIaaS (KI#19) Sensing: Native ISAC (KI#20) Computing: Integrated (KI#22) NTN/Satellite: Native from Day 1 (KI#23) Slicing: Enhanced (KI#3) Intent-based: Yes, with AI agents (KI#18) Spec: TR 23.801 (study) → Rel-21 normative
Figure 1: 5G Core (TS 23.501) vs 6G Core (TR 23.801) Architecture — Side-by-side comparison showing all Key Issue references

Architectural Assumptions (TR 23.801 Section 4.1)

The document establishes 5 foundational assumptions that frame everything:

1. SBA is the starting point
The Service-Based Architecture from 5GC is not discarded — it is evolved. 6G will enhance SBA, not replace it.
2. Standalone 6G architecture
6G RAN connects to a single core type: 6G CN. No NSA (Non-Standalone) like early 5G. Clean standalone from day one.
3. RAN/CN split maintained
The separation between RAN and Core functionality from 5G is preserved. No merging of RAN and Core roles.
4. IMS for voice/video
Real-time voice and video (MMTel) will continue to use IMS. No new voice architecture — IMS evolves into “Vo6G”.
5. Native TN + NTN support
6G System natively supports both Terrestrial Networks and Non-Terrestrial Networks (satellites, HAPS) from day one. Not bolted-on.

Architectural Requirements (Section 4.2)

Two mandatory (“shall”) requirements: roaming support and multi-vendor interoperable interfaces. Plus design principles: cloud-native, energy efficient, resilient, and minimal NF interdependencies.

3GPP Reference: TR 23.801-01 V0.4.0 (2026-02), “Study on Architecture for 6G System; Stage 2,” Release 20. Sections 4.1 and 4.2.

• • •
Chapter Three

TR 23.801: The Blueprint for 6G Core

TR 23.801 is organized into 8 Work Tasks (WTs), each containing one or more Key Issues (KIs). Think of Work Tasks as departments and Key Issues as the specific problems each department is solving. Here’s the complete map:

Work Task Scope Key Issues
WT#1 Core Architecture (signalling, SBA, slicing, UP, QoS, policy, exposure, sharing, services) KI#1–16 (16 issues!)
WT#2 Migration & Interworking KI#17
WT#3 AI (AI for 6G + 6G for AI) KI#18, KI#19
WT#4 Integrated Sensing & Communication KI#20
WT#5 Data Framework KI#21
WT#6 Computing Support KI#22
WT#7 Non-Terrestrial Networks KI#23
WT#8 Cellular IoT Enablers KI#24
Think of it like this

Imagine 3GPP is renovating a massive building (the core network). WT#1 is the main construction team handling plumbing, wiring, rooms and doors. WT#2 is the team building the bridge to the old building next door (5G). WT#3 installs the AI brain. WT#4 adds radar sensors to every wall. WT#5 builds the data vault. WT#6 puts computers in every room. WT#7 launches satellite dishes on the roof. WT#8 makes sure billions of tiny IoT sensors can plug in.

• • •
Chapter Four

The 24 Key Issues at a Glance

Here is every Key Issue from TR 23.801, mapped to its 5G equivalent (where applicable), with what’s new in 6G highlighted. This is the most complete summary available anywhere.

KI# Title 5G Equivalent What’s New in 6G
1Control Signalling for 6GSNAS (TS 23.502)Modular NAS — add new functions without impacting existing ones
2SBA FrameworkSBA (TS 23.501)Optimized NF discovery, better scalability & load balancing
3Network SlicingNSSF/SlicingSimplified slicing, better app-to-slice mapping
4User Plane ArchitectureUPF (CUPS)Enhanced CP-UP split, flexible UPF selection, resilience
5QoS Framework5QI/QoS FlowsDynamic QoS, AI traffic support, UE/NW QoS collaboration
6Policy & ChargingPCF/CHFSimplified policy, converged associations, UE input for PCC
7Network ExposureNEFCommon framework, UE/app exposure, intent-based exposure
8Network SharingMOCN/GWCNMulti-Operator Core Network + Indirect sharing for 6G
9Localized Service AccessNPN (SNPN/PNI-NPN)Local subscription mgmt, NPN reuse for 6G
10FWA (Fixed Wireless)FWA in 5GAddress 5G deployment issues for efficient FWA
11Non-3GPP AccessN3IWF/TNGFUntrusted access, Wi-Fi offload, service continuity
12Voice Services (Vo6G)VoNR/IMSNative IMS over 6G, voice continuity 6G↔NR, IMS compatibility
13Emergency ServicesNG-eCallEmergency sessions via 6GS, fallback to NR/E-UTRAN
14Location Services5G LCS (LMF)Redesigned architecture, decouple from AMF, enhanced accuracy
15Messaging (SMS)SMS over NASSMS over NAS & SMS over IP in 6GS
16Regulatory ServicesMPS/MCX/PWSMPS, Mission Critical, Public Warning in 6GS
17Migration & InterworkingEPS↔5GS IWK5GS↔6GS migration, standalone + MRSS options
18AI for 6G ArchitectureNWDAF (limited)NEW AI agents, intent processing, closed-loop, operator control
196G Network for AIN/ANEW UE AI agents, AIaaS, distributed inference/training
20ISAC (Sensing)N/ANEW Sensing service, authorization, data collection, exposure
21Data FrameworkN/ANEW Data discovery, collection, labelling, processing, storage
22Computing SupportMEC (separate)NEW Integrated compute, site discovery, service continuity
23NTN SupportNTN (Rel-17 bolt-on)Native satellite/HAPS support, service continuity
24IoT EnablersCIoT/RedCapAnalyse which 5G IoT features to carry into 6G

Key insight: 6 of the 24 issues are entirely NEW — they have no 5G equivalent: AI for 6G (#18), 6G for AI (#19), ISAC sensing (#20), data framework (#21), computing support (#22). This is where 6G Core fundamentally diverges from 5G Core.

• • •
Chapter Five

SBA Evolution, Signalling & User Plane

KI#1: Modular Control Signalling

The biggest pain point in 5G: adding a new NAS feature (like a new mobility procedure) often requires changes to multiple NFs. KI#1 studies how to make NAS modular — introduce new functionalities with minimal or no impact to existing ones. Two sub-issues:

KI#2: SBA Framework Optimization

5G’s SBA works, but at scale it shows strain. KI#2 targets two improvements:

  1. Efficient NF discovery & selection — reduce message overhead for finding the right NF
  2. Better resiliency & load balancing — handle NF failures more gracefully than 5G

KI#4: User Plane Architecture

The UPF (User Plane Function) carries all your actual data. KI#4 enhances it in three ways:

5G User Plane

  • UPF selected by SMF
  • Single CP-UP split model
  • Limited multi-vendor UP interop
  • Basic resilience mechanisms

6G User Plane

  • Flexible UPF (re-)selection based on capability & path performance
  • Enhanced CP-UP split for better multi-vendor interop
  • Built-in high availability
  • UPF + Computing Site integration (KI#22)
• • •
Chapter Six

AI in 6G Core: Agents, Intents & AIaaS

This is the most transformative part of TR 23.801. Two Key Issues define how AI enters the 6G Core:

KI#18: AI for 6G Architecture (The Core Gets Smart)

KI#18 is about putting AI inside the core network to make it self-managing. It introduces the concept of AI agents — autonomous entities that perform tasks on behalf of operators. Nine specific aspects are being studied:

1. Intent Processing
AI agents process “intents” from UEs and AFs. An intent is an expectation (“ensure 99.999% reliability”) without specifying how. The AI figures out the implementation. Intents must be unambiguous and interoperable.
2. Dynamic Procedure Composition
AI agents can dynamically compose procedures from modular building blocks (linked to KI#1). Instead of hard-coded call flows, the AI assembles the right sequence of steps based on the situation.
3. Closed-Loop Operations
Support for reinforcement learning and closed-loop automation. The network monitors outcomes, learns, and adjusts. No human in the loop for routine optimization.
4. Operator Control & Autonomy Levels
Operators configure how much autonomy the AI has. From “AI suggests, human approves” to “AI acts autonomously.” Including the option to not use AI at all.
5. AI Performance Monitoring
All AI entities must be monitorable. If an AI agent starts making bad decisions, the operator can see it, stop it, or roll back. Trustworthiness built-in.

Critical design principle (from TR 23.801): “It is assumed it is not required for the MT stack of UE to produce nor understand the intent. The MT stack of UE is assumed to be agnostic to whether or not the network uses AI capable entities.” — In other words: your phone doesn’t need to know the network is using AI. It just works.

KI#19: 6G Network for AI (The Network Serves AI)

This is the reverse: instead of AI helping the network, the network helps AI applications. Five aspects:

1. UE AI Agent Discovery
AI agents on different UEs can discover each other via the 6G network. Your phone’s AI assistant could find and communicate with a robot’s AI agent.
2. AI Agent Communication
Secure, identified, authorized communication between AI agents on different devices via the network.
3. Network Capability Exposure for AI
Expose network capabilities to AI agents running on application servers (AFs). Coordinated with KI#7.
4. AI-as-a-Service (AIaaS)
The 6G CN provides AI inferencing and AI training services to applications and UEs via the network.
5. AI Traffic Impact
Study system-level impacts of AI-specific traffic patterns (bursty, latency-sensitive, large model transfers).
Think of it like this

KI#18 is like putting an AI brain inside the hospital to manage operations. KI#19 is like the hospital offering AI diagnostic services to external clinics — any doctor can send a scan and get back an AI analysis.

• • •
Chapter Seven

Sensing (ISAC) & Computing Support

KI#20: Integrated Sensing and Communication

This is entirely new — nothing like it exists in 5G. The 6G network will use its radio signals not just for data, but as radar to sense the physical environment. TR 23.801 studies the core network architecture needed to support this:

KI#22: 6G Computing Support

In 5G, MEC (Multi-access Edge Computing) is a separate platform bolted onto the core. In 6G, computing is integrated into the architecture. KI#22 studies:

5G: MEC (Separate)

  • MEC is an ETSI standard, separate from 3GPP
  • Deployed alongside the UPF
  • Limited integration with SMF/PCF
  • No native compute discovery
  • No compute-aware session management

6G: Integrated Compute

  • Computing Sites are first-class 3GPP entities
  • Compute resource discovery & (re-)selection
  • Communication + Compute coordinated
  • Service continuity when compute site changes
  • Compute-aware QoS (KI#5 coordination)

KI#21: Data Framework

With AI, sensing, and computing all generating massive amounts of data, 6G needs a unified data framework. KI#21 covers: data discovery & registration, collection & transfer, labelling, processing (anonymization, analysis), storage, and exposure. All with privacy, consent, and governance baked in.

• • •
Chapter Eight

Network Slicing 2.0 & QoS Framework

KI#3: Enhanced Slicing

Network slicing was 5G’s killer feature. In 6G, KI#3 focuses on simplification and improvement:

KI#5: QoS Framework for 6G

The QoS framework gets the most extensive study with 8 solution variants already proposed. Key enhancements:

Dynamic QoS Requirements
Applications can change their QoS needs in real-time. A video call might switch from standard to ultra-low-latency when the user starts a surgical procedure.
AI Traffic Support
New QoS handling for AI/ML application traffic which has unique characteristics: bursty inference requests, large model uploads, latency-sensitive responses.
UE/Network QoS Collaboration
The UE and network share awareness: the network tells the app what it can provide, the app tells the network what it needs. Two-way negotiation instead of one-way enforcement.
Graceful Degradation
When the network can’t meet QoS targets, instead of just dropping quality, it negotiates adjusted targets to minimize impact on user experience.
• • •
Chapter Nine

Migration & Interworking (5G ↔ 6G)

KI#17: The Bridge Between Generations

No operator will switch to 6G overnight. KI#17 ensures a smooth transition:

Migration to 6GS
How operators gradually transition from 5G to 6G infrastructure. Multiple migration options being studied: standalone 6G, Multi-RAT Spectrum Sharing (MRSS), and inter-RAT mobility between NR and 6GR.
Interworking with 5GS
How 5G and 6G systems work side by side. A UE should seamlessly move between 5G and 6G coverage. This works even if the UE was previously in 2G, 3G, or 4G.
NTN Interworking
How 6GS interworks with 4G/5G satellite access already in orbit. You can’t replace all satellites at once — the systems must coexist.

Note: Interworking with 2G/3G is not being studied for 6G (TR 23.801 explicitly excludes it). The assumption is that by 2031, 2G/3G networks will be sunset in most markets.

• • •
Chapter Ten

Voice, Emergency & Essential Services

These “boring” services are actually critical — voice and emergency calls must work perfectly from day one of 6G launch.

KI#12: Vo6G (Voice over 6G)

Voice will use IMS (IP Multimedia Subsystem), just like VoNR in 5G. But there are real challenges:

KI#13 & KI#14: Emergency & Location

Emergency calls over 6G must work even if the UE is in limited-service state. Location services are being redesigned — one of the most interesting architectural changes is decoupling the location function from the AMF (KI#14, Solution #14.2), allowing a dedicated 6G-LMF with direct RAN connection. This enables more accurate and faster positioning.

KI#7: Network Exposure & Intents

6G introduces intent-based exposure on the northbound interface. Instead of an application asking for specific QoS parameters, it expresses an intent: “I need low-latency, high-reliability communication for this surgical robot.” The 6G CN interprets and fulfills it. This bridges KI#7 with KI#18 (AI).

• • •
Chapter Eleven

Preparing for the 6G Core Era

For Core Network Engineers

Master 5G Core (TS 23.501/502/503)
6G Core evolves from 5GC. Every 6G key issue uses 5G as the “starting point for discussion.” If you don’t deeply understand AMF, SMF, UPF, PCF, NRF, NSSF, NEF — start here.
Learn AI/ML Fundamentals
KI#18 and KI#19 make AI central to 6G Core. Understand reinforcement learning, intent processing, and ML model lifecycle.
Follow TR 23.801 Updates
The study is ongoing. Subscribe to 3GPP SA2 meeting reports. Watch for solution evaluations and interim agreements in Section 7.
Experiment with Open-Source Cores
Platforms like Open5GS, free5GC, and Aether provide hands-on SBA experience. The patterns you learn (NF registration, service discovery, API-based communication) carry directly into 6G.

“The study will work towards goals to create lean and streamlined standards for 6G, minimizing the adoption of multiple options for the same functionality, avoiding excessive configurations.”

— TR 23.801 Introduction, endorsed at TSG#107 (March 2025)

6G Core isn’t a distant dream — it’s being designed right now in TR 23.801. The 24 key issues and 8 work tasks we’ve decoded here will become the normative specifications of 3GPP Release 21. Understanding them today puts you years ahead of the curve.

Master 5G Core & Prepare for 6G

CafeTele offers comprehensive courses on 5G Core architecture, physical layer, and AI in telecom.