27.02.2026 - New Project: 6G Testbed & AI Traffic Characterization

18 hours ago
1 min read

A perfect way to end the week. Join us on Friday 27th February at 13:00 CET for a deep dive into the new 5G-MAG Reference Tools Project "6G Testbed and AI Traffic Characterization".

This is a new open-source project maintained by 5G-MAG which provides a specialized environment for testing and characterizing how Artificial Intelligence (AI) applications interact with next-generation networks.

Linked to the new 3GPP SA4 Study on 6G Media, the projects aims at helping the industry understand and evaluate how different AI engines consume network resources and tolerate network issues.

Key Technical Components

Scenario Engine: Supports over 11 scenario types, including:
- Agentic AI: Using Model Context Protocol (MCP) tools.
- Multimodal: Handling simultaneous audio, video, and image data.
- Real-time: Testing low-latency variants using WebRTC or WebSockets (e.g., OpenAI Realtime).
Network Emulator: A Linux-based tool that simulates various 3GPP-compliant network conditions:
- Profiles: Includes pre-defined settings for Ideal 6G, 5G Urban, Cell Edge, Satellite (NTN), and Congested environments.
- Controls: Allows precise manipulation of delay, loss, and bandwidth shaping.
LLM Provider Integration: Interfaces directly with major AI providers like OpenAI, Gemini, DeepSeek, and vLLM to generate realistic application traffic.
Metric Collection: Measures over 60 KPIs, focusing on metrics critical to AI performance, such as:
- TTFT (Time to First Token) and TPOT (Time Per Output Token).
- Latency percentiles and Uplink/Downlink ratios.
- Agent loop factors (how network quality affects an AI's ability to complete multi-step tasks).