1. Telecom-Specific Foundation Models

Unlike generic models (like GPT-4), these open-source projects are training AI specifically to understand complex 3GPP standards, network logs, and configuration protocols.

• GSMA "Open-Telco" LLM Benchmarks & Models:

  • What it is: A major industry initiative (hosted on Hugging Face) released in late 2024/2025 to standardize how AI is used in telecom.

  • Key Tool: TeleQnA. This is an open-source dataset and evaluation suite. It allows developers to test if their AI model actually understands telecommunications (e.g., "How do I fix a handover failure in 5G SA?") rather than just writing generic code.

  • Development: In 2025, they released "Telco-RAG", a retrieval-augmented generation framework that allows operators to safely connect LLMs to their private technical manuals.

• Anlix (Log Analysis):

  • What it is: An open-source tool for analyzing massive streams of network logs.

  • Capability: It uses machine learning to detect anomalies in ISP networks. Instead of an engineer manually grepping through logs, Anlix clusters error messages to find the "root cause" of a WiFi outage instantly.

2. 6G Research & Physical Layer AI

These tools are used to design the "Air Interface" of the future, where AI (not human math) decides how to encode data onto radio waves.

• NVIDIA Sionna:

  • What it is: The industry-standard open-source library (based on TensorFlow/JAX) for 6G physical layer research.

  • 2025 Update: Recent updates have added "Differentiable Ray Tracing." This allows researchers to create a "Digital Twin" of a city (e.g., downtown Berlin) and train an AI to predict exactly how radio waves will bounce off buildings.

  • Use Case: Engineers use Sionna to train "Neural Receivers"—AI blocks that replace traditional signal processing components (like the FFT or channel estimator) to get faster speeds in noisy environments.

3. O-RAN (Open Radio Access Network) AI

The O-RAN Alliance is the primary driver of open AI in telecom. Their architecture splits the base station into pieces, allowing AI "apps" to control the radio.

• O-RAN Software Community (O-RAN SC) - RIC Platform:

  • What it is: The open-source code for the RAN Intelligent Controller (RIC).

  • The "App Store" Concept:

    • Near-RT RIC (Real-Time): Hosts "xApps". These are open-source AI scripts that make decisions in <10ms (e.g., identifying jamming and switching frequencies instantly).

    • Non-RT RIC (Non-Real-Time): Hosts "rApps". These run in the cloud and handle long-term optimization (e.g., putting cell towers to sleep at night to save power).

• OSC AI/ML Framework:

  • Development: In 2025, the O-RAN SC released a standardized "AI/ML Workflow" that allows models trained in the cloud (using tools like Kubeflow) to be automatically packaged and deployed to the cell tower edge as xApps.

4. Network Orchestration & Automation

Tools that manage the "Cloud" side of the telecom network.

• ETSI Open Source MANO (OSM) Release 17/18:

  • What it is: The standard open-source stack for Management and Orchestration.

  • AI Feature: The latest releases focus on "Closed-Loop Automation."

  • Example: If the AI detects that a specific "Network Slice" (e.g., a slice dedicated to emergency services) is getting congested, OSM can automatically spin up more virtual machines to handle the load without a human operator approving it.

• Magma Core (Linux Foundation):

  • What it is: An open-source mobile packet core.

  • Capability: While primarily a core network, recent community extensions allow for "Federated Learning." This enables multiple private 5G networks (e.g., in different factories) to train a shared security model without sharing their sensitive raw data.

1. Network Automation & Management

Network Orchestration Platforms

• ONAP (Open Network Automation Platform) - Linux Foundation project

  • Source: https://github.com/onap

  • Features: Network service orchestration, closed-loop automation, AI/ML workflows

  • Components: DCAE (Data Collection, Analytics, Events), Policy Framework, CLAMP (Closed Loop Automation Management Platform)

• Open Source MANO (OSM) - ETSI-hosted project

  • Source: https://osm.etsi.org/gitlab/

  • Features: NFV Management and Orchestration, includes monitoring and analytics capabilities

• OpenDaylight - SDN controller platform

  • Source: https://github.com/opendaylight

  • Features: Modular SDN controller, network programmability, plugins for analytics

• Tungsten Fabric - Multi-cloud networking platform

  • Source: https://github.com/tungstenfabric

  • Features: SDN controller, analytics, security policy enforcement

Network Analytics & Monitoring

• SNAS (Streaming Network Analytics System) - Real-time network analytics

  • Source: https://github.com/IntelLabs/SNAS

  • Features: Real-time streaming analytics for network traffic, anomaly detection

• PNDA (Platform for Network Data Analytics) - Big data analytics platform for networks

  • Source: https://github.com/pndaproject

  • Features: Collects and analyzes network data, ML-based analytics pipelines

• Apache Spot (incubating) - Cybersecurity analytics platform

  • Source: https://github.com/apache/incubator-spot

  • Features: Network traffic analysis, ML-based threat detection

• Grafana & Prometheus - Monitoring stack with ML capabilities

  • Sources: https://github.com/grafana/grafana, https://github.com/prometheus/prometheus

  • Features: Time-series monitoring, alerting, integration with ML frameworks

2. Radio Access Network (RAN) & 5G

Open RAN Projects

• O-RAN SC (Software Community) - Open source implementations for O-RAN

  • Source: https://github.com/o-ran-sc

  • Key Projects:

    • RIC (RAN Intelligent Controller) platform

    • Non-RT RIC - For non-real-time control and AI/ML applications

    • Near-RT RIC - For near-real-time control

    • xApps and rApps - Applications for RIC

• OpenAirInterface (OAI) - Open-source 5G implementation

  • Source: https://gitlab.eurecom.fr/oai

  • Features: Full 5G stack implementation, includes AI/ML integration points

• srsRAN - Open-source 4G/5G RAN implementation

  • Source: https://github.com/srsran/srsRAN

  • Features: Complete LTE and 5G NR stack, extensible for AI/ML

• OpenRAN Gym - AI/ML research platform for Open RAN

  • Source: https://github.com/opengym/OpenRAN-gym

  • Features: Simulator and dataset for developing AI for RAN optimization

5G Core & Edge

• Free5GC - Open-source 5G Core Network

  • Source: https://github.com/free5gc/free5gc

  • Features: 3GPP-compliant 5G core, extensible for AI applications

• Open5GS - Open-source 5G Core and EPC

  • Source: https://github.com/open5gs/open5gs

  • Features: 4G/5G core network implementation

• Magma - Open-source wireless core (now part of Linux Foundation)

  • Source: https://github.com/magma/magma

  • Features: Mobile core for 4G/5G, includes analytics framework

3. Network Simulation & Testing

• ns-3 - Discrete-event network simulator

  • Source: https://gitlab.com/nsnam/ns-3-dev

  • Features: Large-scale network simulation, supports ML integration via Python bindings

• OMNeT++ - Component-based network simulation

  • Source: https://github.com/omnetpp/omnetpp

  • Features: Modular simulation framework, INET Framework for Internet protocols

• Mininet - Network emulator for SDN

  • Source: https://github.com/mininet/mininet

  • Features: Creates virtual networks on a single machine, great for testing AI-based SDN controllers

• ONOS (Open Network Operating System) - SDN controller

  • Source: https://github.com/opennetworkinglab/onos

  • Features: Carrier-grade SDN controller, includes ML-based applications

4. AI/ML Frameworks for Telecom

Specialized ML Libraries

• NetworkX - Python library for network analysis

  • Source: https://github.com/networkx/networkx

  • Features: Network/graph analysis, topology analysis, graph ML algorithms

• Karate Club - Network embedding and community detection

  • Source: https://github.com/benedekrozemberczki/karateclub

  • Features: Graph mining, network representation learning

• Deep Graph Library (DGL) - Graph neural networks

  • Source: https://github.com/dmlc/dgl

  • Features: Graph neural networks for network analysis and optimization

Time Series Analysis for Network Data

• Merlion - Time series intelligence library

  • Source: https://github.com/salesforce/Merlion

  • Features: Anomaly detection, forecasting for network metrics

• Prophet - Forecasting for time series data

  • Source: https://github.com/facebook/prophet

  • Features: Automatic forecasting of network traffic, capacity planning

• Orbit - Bayesian time series forecasting

  • Source: https://github.com/uber/orbit

  • Features: Probabilistic forecasting for network planning

5. Security & Threat Intelligence

• Zeek (formerly Bro) - Network security monitor

  • Source: https://github.com/zeek/zeek

  • Features: Network traffic analysis, ML-based intrusion detection extensions

• Suricata - Network IDS/IPS

  • Source: https://github.com/OISF/suricata

  • Features: Real-time intrusion detection, ML capabilities via Lua scripts

• Moloch - Network traffic analysis

  • Source: https://github.com/aol/moloch

  • Features: Large-scale traffic capture and indexing, extensible for ML

• Apache Metron (incubating) - Security analytics platform

  • Source: https://github.com/apache/metron

  • Features: Security event processing, ML-based threat scoring

• OpenCTI - Cyber threat intelligence platform

  • Source: https://github.com/OpenCTI-Platform/opencti

  • Features: Threat intelligence management, ML for threat correlation

6. Data Platforms & Analytics

• Apache Druid - Real-time analytics database

  • Source: https://github.com/apache/druid

  • Features: High-performance time-series database for network metrics

• Apache Flink - Stream processing framework

  • Source: https://github.com/apache/flink

  • Features: Real-time stream processing for network data, ML integration

• Apache Kafka - Distributed streaming platform

  • Source: https://github.com/apache/kafka

  • Features: Real-time data pipelines for network telemetry

• ClickHouse - Column-oriented database for analytics

  • Source: https://github.com/ClickHouse/ClickHouse

  • Features: High-performance analytics for network data

7. Network Testing & Validation

• Bess (Berkeley Extensible Software Switch) - Software switch for testing

  • Source: https://github.com/NetSys/bess

  • Features: Modular software switch, used for testing ML-based traffic management

• P4 - Programming protocol-independent packet processors

  • Source: https://github.com/p4lang

  • Features: Language for programming network devices, can integrate with ML

• CORE (Common Open Research Emulator) - Network emulator

  • Source: https://github.com/coreemu/core

  • Features: Real-time network emulation for testing AI algorithms

8. MLOps & Model Management

• MLflow - Machine learning lifecycle platform

  • Source: https://github.com/mlflow/mlflow

  • Features: Experiment tracking, model deployment, model registry

• Kubeflow - Machine learning on Kubernetes

  • Source: https://github.com/kubeflow/kubeflow

  • Features: End-to-end ML workflows, deployment of ML models in telecom networks

• Seldon Core - ML model deployment on Kubernetes

  • Source: https://github.com/SeldonIO/seldon-core

  • Features: Model serving, monitoring, explainability for telecom AI models

• BentoML - Model serving framework

  • Source: https://github.com/bentoml/BentoML

  • Features: High-performance model serving, ideal for network inference

9. Research & Academic Projects

University Research

• AutoML for Networks - MIT's AutoML for network configuration

  • Source: Various research codebases from networking labs

  • Focus: Automatic optimization of network parameters using ML

• RouteNet - GNN-based network performance predictor

  • Source: https://github.com/knowledgedefinednetworking/RouteNet

  • Features: Graph neural networks for network performance modeling

• Knowledge-Defined Networking (KDN) - Framework from UPC Barcelona

  • Source: https://github.com/knowledgedefinednetworking

  • Projects: Various tools for ML in networking

Industry Research Labs

• Facebook's Network AI Research - Various open-source contributions

  • Examples: Katran (load balancer), Open/R (routing protocol)

• Google's BBR (Bottleneck Bandwidth and Round-trip) - Congestion control algorithm

  • Source: Implemented in Linux kernel

• Microsoft's SONiC (Software for Open Networking in the Cloud)

  • Source: https://github.com/Azure/SONiC

  • Features: Network operating system with telemetry and analytics

10. Standards-Based Implementations

3GPP NWDAF Implementations

• OpenNWDAF - Open-source implementation of 3GPP NWDAF

  • Source: Various research implementations (e.g., from universities)

  • Features: Network Data Analytics Function for 5G, analytics exposure

ETSI ENI

• OpenENI - Reference implementations of ETSI ENI

  • Source: ETSI provides reference implementations

  • Features: Experiential Networked Intelligence architecture

11. Visualization & Dashboarding

• Grafana Network Panel Plugins - Specialized plugins for network visualization

  • Various plugins: For network topology, flow visualization, etc.

• NetBox - IP address management and data center infrastructure

  • Source: https://github.com/netbox-community/netbox

  • Features: Network infrastructure management, extensible with plugins

12. Telecom-Specific Datasets

• MAWI Working Group Traffic Archive - Internet traffic traces

  • Source: https://mawi.wide.ad.jp/mawi/

• CAIDA Datasets - Internet measurement data

  • Source: https://www.caida.org/catalog/datasets/

• UMass Trace Repository - Network traces

  • Source: https://traces.cs.umass.edu/