Stella EN 1.5B V5
NovaSearch/stella_en_1.5B_v5
published Jul 2024 · updated Jul 2025
Stella EN 1.5B V5 is an embedding model that transforms text into numerical vectors for dense retrieval, supporting multiple output dimensions via Matryoshka Representation Learning.
specs
| Task | Embedding (dense retrieval, semantic similarity) |
| Architecture | Based on Alibaba-NLP/gte-Qwen2-1.5B-instruct |
| Parameters | 1.5B |
| License | MIT |
best for
- ·Retrieval-Augmented Generation (RAG) document retrieval
- ·Semantic textual similarity (STS) tasks
- ·FAQ matching and question-answering systems
FAQ
It supports 512, 768, 1024, 2048, 4096, 6144, and 8192 dimensions. 1024 is the default and recommended for most use cases.
Use the "s2p_query" prompt for sentence-to-passage retrieval tasks and the "s2s_query" prompt for sentence-to-sentence semantic similarity tasks.
The model is trained on sequences of 512 tokens; longer sequences may degrade performance.
Use the gigarouter OpenAI-compatible endpoint with your API key, specifying the model name "NovaSearch/stella_en_1.5B_v5".
It is released under the MIT license.
# OpenAI client - just change base_url from openai import OpenAI client = OpenAI(base_url="https://gigarouter.ai/v1", api_key=KEY) v = client.embeddings.create(model="NovaSearch/stella_en_1.5B_v5", input=["hello world"]) print(v.data[0].embedding[:4])
try it live
runs the real hosted model on a shared demo allowance · get your own key + $25 free →