AI Agents for RAN Challenge by ITU
AI Agents for RAN Challenge by ITU supports participants building AI agents to solve radio network engineering tasks through simulation, advancing reasoning and tool use for more autonomous network operations.
Funder: Zindi
Due Dates: November 27, 2026 (Start) | February 5, 2027 (Submission close) | February 5, 2027 (Submission selection close) | February 6, 2027 (Leaderboard reveal)
Funding Amounts: €20,000 in prizes; top three also receive MWC Barcelona travel grants up to €3,500 each
Summary: An ITU challenge hosted on Zindi for AI agents solving radio network engineering tasks through simulation and structured reasoning.
Key Information: All agents must use Qwen3.6-35B-A3B as their base language model.
Description
The AI Agents for RAN Challenge by ITU, hosted on Zindi, invites participants to build agents that solve radio network engineering problems using Giulia, an interactive network simulator. It tests reasoning across complex radio access networks, where changing one site or cell can affect coverage, interference, user association, throughput and energy consumption elsewhere.
Agents receive a scenario and objective, use dedicated tools to inspect and simulate the network, and produce a structured answer. Tasks cover fault diagnosis, outage impact analysis, coverage, radio configuration, site placement and energy management. Some require analytical answers; others require configurations meeting specified performance conditions. Each task limits simulations, making experiment selection central to the challenge. Testing on unfamiliar scenarios with measurable outcomes aims to advance reasoning and tool use for more autonomous network operations.
