# NetoAI

> NetoAI builds sovereign, telecom-specific AI for autonomous networks. Creator of TSLAM — the industry's first telecom-specific LLM, independently benchmarked by the GSMA (topped the TeleQnA domain-reasoning leaderboard in the GSMA Open-Telco LLM Benchmarks 2.0) — and TelcoCore, an AI-native product suite spanning agentic AI (ViNG), a graph-native network digital twin (Digi-Twin) and network-as-an-API orchestration (NAPI). ISO/IEC 42001:2023 and ISO/IEC 27001:2022 certified. Deploys on-premise, air-gapped, at the edge, or in the cloud. Offices in the UK and India.

Key facts:
- TSLAM-G3 topped the GSMA Open-Telco LLM Benchmarks 2.0 domain-reasoning (TeleQnA) leaderboard with a score of 82.5, among 23 models including the largest frontier labs.
- TSLAM-18B ranked #3 in TeleMath and 3GPP-TSG (network engineering and standards), within 5% of the top 600B+ generalist model while being ~30× smaller.
- TSLAM is trained on 100B+ tokens of telecom-specific data and runs fully sovereign (on-premise / air-gapped / edge).
- Open-source models (TSLAM, T-VEC, NetBench) have 26,000+ downloads on Hugging Face: https://huggingface.co/NetoAISolutions
- NetoAI built the GSMA Agentic AI Testbed, a neutral sandbox for GSMA member operators to evaluate agentic AI on live network scenarios.
- TM Forum ODA Component Directory partner; AWS partner (SageMaker + Trainium); research published at EMNLP 2025 and on arXiv.

## Products

- [TSLAM — Telecom-Specific LLM](https://netoai.ai/tslam/): The industry's first telecom LLM family (open-source 2B/4B/8B; enterprise 18B/30B/108B; CxLAM for customer experience), plus T-VEC telecom embeddings and T-Synth sub-600ms voice.
- [ViNG — Agentic AI Platform](https://netoai.ai/ving/): Low-code/no-code multi-agent platform powered by TSLAM. Supports A2A and Model Context Protocol (MCP); 56+ enterprise connectors.
- [ViNG for Engineers](https://netoai.ai/ving-for-engineers/): Free AI workbench for NOC, RF and RAN engineers — paste KPIs, alarms, logs or tickets and get likely causes, next checks and RCA-ready summaries. Early access waitlist.
- [Digi-Twin — Network AI Brain](https://netoai.ai/digi-twin/): Real-time, graph-native (bitemporal) digital twin of the network for simulation, root-cause analysis and predictive operations.
- [NAPI — Network Orchestrator](https://netoai.ai/napi/): Turns legacy and modern networks into a unified, TM Forum-compliant API (TMF638/639/640). AI-driven device integration in ~2 minutes; 84 pre-built connectors.

## Proof & Research

- [GSMA benchmark results](https://netoai.ai/benchmarks/): Full GSMA Open-Telco LLM Benchmarks 2.0 results for TSLAM, with methodology and sources.
- [Resources](https://netoai.ai/resources/): Peer-reviewed papers (EMNLP 2025, arXiv), GSMA co-authored whitepapers, AWS partner blog, Everest Group analyst review, open-source models.

## Company

- [About NetoAI](https://netoai.ai/company/): Founded 2024. Mission: sovereign, telecom-specific AI that takes operators to Level 4 autonomous networks (TM Forum framework).
- [Careers](https://netoai.ai/careers/): Hiring AI researchers, ML engineers and builders.
- [Contact](https://netoai.ai/#contact): Schedule a demo, or email info@netoai.ai.