OTel-Embedding-4B

OTel-Embedding-4B is a telecom embedding model full-parameter fine-tuned on OTel telecommunications data. It is part of the OTel Family of Models, an open-source initiative to build reference AI resources for the global telecommunications sector.

Across the OTel embedding baselines, OTel fine-tuning improves telecom retrieval quality by +9.6 to +60.2 NDCG@10 percentage points.

Community Use

As of June 23, 2026, the released OTel models had more than 18 million downloads, and the Open Telco AI project had received 157+ pieces of media coverage worldwide.

Model Details

Attribute	Value
Base model	Qwen/Qwen3-Embedding-4B
Parameters	4B
OTel training dataset	OTel-Embedding
Dataset fields	`anchor`, `positive`, `negative_1` through `negative_5`
Training method	Full-parameter post-training / fine-tuning
Language	English
OTel release license	Apache 2.0

Model Lineage

Qwen/Qwen3-Embedding-4B -> OTel-Embedding full-parameter post-training -> farbodtavakkoli/OTel-Embedding-4B

OTel vs. Base Model

Metric	Base model	OTel fine-tuned	Delta	Evaluation split
NDCG@10	82.5%	92.2% +/- 0.4	+9.7 pp	OTel-Embedding held-out 10%

Standard errors are computed with bootstrap resampling (n=10) over the held-out OTel evaluation partition. NDCG@10 measures whether relevant telecom passages are ranked near the top of the retrieved list.

Evaluation Caveats

Embedding results measure held-out OTel retrieval partitions.
Reported standard errors come from bootstrap resampling over the held-out evaluation partitions.
Aggregate NDCG@10 does not guarantee uniform performance across all telecom subdomains.
External benchmark transfer, multilingual performance, and per-subdomain performance should be evaluated separately for production settings.

Training Data

The model was trained on telecom-focused data curated by 100+ domain experts. The raw corpus contained roughly 1.1M training points and was filtered to 326,767 higher-confidence examples.

Source	Contributor
arXiv telecom papers, 3GPP standards, telecom Wikipedia, telecom Common Crawl	Yale University
GSMA Permanent Reference Documents, Discover portal	GSMA
IETF RFC series	NetoAI
Industry whitepapers	Khalifa University
O-RAN specifications (working groups 1, 2, 4, 5, 6, 7, 8, 9, 10)	University of Leeds
O-RAN documents across working groups	The University of Texas at Dallas

Released datasets: OTel-LLM, OTel-Embedding, OTel-Reranker, and OTel-Safety.

The OTel datasets release derived QA/retrieval/reranking examples rather than the raw source documents.

Each released dataset includes a dataset card and Croissant metadata with Responsible AI fields for data limitations, biases, sensitive-information considerations, use cases, social impact, synthetic-data status, and provenance.

Representative Training Row

OTel-Embedding rows provide one anchor query, one positive passage, and five hard negatives for telecom retrieval training.

{
  "anchor": "During the Measurement ID Coordination (MN initiated) test case, what is the relationship between the U-Plane data and the F1 logs as observed in the Protocol Analyzer and the Test UE or UE emulator?",
  "positive": "F1 logs recorded in the Protocol Analyzer and the Test UE or UE emulator show that all downlink U-Plane data recorded in the F1 logs is correctly received, and all uplink U-Plane data transmitted by the Test UE or emulated UE is recorded in the F1 logs.",
  "negative_count": 5
}

Intended Use

This model is intended to retrieve relevant chunks from telecom specifications, standards, RFCs, whitepapers, and other domain documents. It can be used as the retrieval stage in an end-to-end telecom RAG pipeline, followed by a reranker and a context-grounded LLM.

Training Recipe

Item	Value
Framework	ScalarLM
Optimizer	AdamW, 8-bit
Learning-rate schedule	Cosine decay with warmup
Weight decay	0.01
Warmup steps	100
Random seed	42
Maximum sequence length	1500 tokens
Precision	BF16
Attention	Flash Attention 2
Distributed training	Fully Sharded Data Parallel
Gradient checkpointing	Enabled
Epochs	3 for LLM/embedding models; 2 for rerankers
Compute	AMD MI300X/MI325X/MI355X and NVIDIA A100/H100 GPUs

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("farbodtavakkoli/OTel-Embedding-4B")

sentences = [
    "What is the F1 interface in O-RAN?",
    "The F1 interface connects the O-RAN Distributed Unit (O-DU) to the O-RAN Central Unit (O-CU).",
]

embeddings = model.encode(sentences, normalize_embeddings=True)
print(embeddings.shape)

Limitations and Responsible Use

OTel models are domain-specific to telecommunications and should not be treated as general-purpose models.
The current release is English-only and primarily text-centric.
The reported OTel performance results use held-out OTel evaluation partitions and should not be interpreted as results from a fully independent external benchmark suite.
Aggregate scores can hide subdomain variation; collaborator stress tests suggest O-RAN retrieval is comparatively strong, while academic-paper and GSMA PRD examples need further curation.
Generated telecom content should be verified before operational, customer-facing, regulatory, safety, or network-configuration use.
Users must comply with both the OTel release license and the upstream base-model license or terms.

Related Models

Project Resources

Project page: https://huggingface.co/farbodtavakkoli
Code: https://github.com/farbodtavakkoli/OTel
Media coverage list: https://github.com/farbodtavakkoli/OTel/blob/main/docs/media_coverage.md

Citation

@misc{otel_models_2026,
  title  = {OTel: Open Telco AI Datasets, Benchmarks, and Models},
  author = {Tavakkoli, Farbod and others},
  year   = {2026},
  note   = {Open Telco (OTel) model release},
  url    = {https://huggingface.co/farbodtavakkoli}
}