Views
No views yet
jinaai/jina-embeddings-v5-text-nano (EuroBERT-210M, 12L/768d) via layer pruning + vocabulary pruning + knowledge distillation with conversation data.| Property | Value |
|---|---|
| Teacher | jinaai/jina-embeddings-v5-text-nano (239M params) |
| Architecture | EuroBERT (pruned) |
| Hidden dim | 256 |
| Layers | 6 / 12 |
| Intermediate | 1024 |
| Model size (FP32) | 64.8 MB |
| Embedding dim | 256 |
| Compression | ~3.7x (239M -> ~16M params) |
1from sentence_transformers import SentenceTransformer
2
3model = SentenceTransformer("gomyk/jina-v5-h256-distilled-conv", trust_remote_code=True)
4embeddings = model.encode(["Hello world", "How are you?"])
5print(embeddings.shape) # (2, 256)jinaai/jina-embeddings-v5-text-nano (12 layers, 768d, 128K vocab)| Task Group | Average |
|---|---|
| Classification | 63.86% |
| Clustering | 32.62% |
| STS | 70.60% |
| Task | Score |
|---|---|
| AmazonCounterfactualClassification | 72.81% |
| Banking77Classification | 77.90% |
| ImdbClassification | 73.39% |
| MTOPDomainClassification | 86.91% |
| MassiveIntentClassification | 32.96% |
| MassiveScenarioClassification | 40.19% |
| ToxicConversationsClassification | 62.45% |
| TweetSentimentExtractionClassification | 64.25% |
| Task | Score |
|---|---|
| ArXivHierarchicalClusteringP2P | 48.36% |
| ArXivHierarchicalClusteringS2S | 46.85% |
| BiorxivClusteringP2P.v2 | 18.36% |
| MedrxivClusteringP2P.v2 | 24.59% |
| MedrxivClusteringS2S.v2 | 23.35% |
| StackExchangeClustering.v2 | 42.75% |
| StackExchangeClusteringP2P.v2 | 33.83% |
| TwentyNewsgroupsClustering.v2 | 22.86% |
| Task | Score |
|---|---|
| BIOSSES | 63.87% |
| SICK-R | 81.44% |
| STS12 | 75.81% |
| STS13 | 76.53% |
| STS14 | 77.58% |
| STS15 | 84.28% |
| STS17 | 57.45% |
| STS22.v2 | 34.21% |
| STSBenchmark | 84.20% |