We introduce Open Telco (OTel), an open resource for telecom AI. OTel releases derived telecom datasets for retrieval, reranking, instruction tuning, and safety/abstention, together with 30 full‑parameter post‑trained baselines: 10 embedding models, 3 rerankers, and 17 language models. As of May 3, 2026, the models have been downloaded over 16 million times and received more than 157 media mentions worldwide.
Building on existing open telecom datasets and benchmarks, OTel consolidates documented data sources, held‑out evaluation splits, trained embedding models, rerankers, context‑grounded LLMs, and safety/abstention data in a single package. Each baseline starts from an open‑weight model, follows a publicly released training recipe on OTel‑derived data, and is evaluated on the held‑out splits. Post‑training improves performance across all three families: retrieval reaches 93.1% NDCG@10, reranking attains 0.947 MRR@10, and language‑model correctness climbs to 87.8%. We release OTel as a reproducible starting point and invite the community to expand the data, enhance embedding and reranking models, and build stronger context‑grounded telecom LLMs.
Review