Model reference · open weights
rubert-large-uncased-sts is an open-weight embedding model from sergeyzh. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.
Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.
What it is
| Released by | sergeyzh |
|---|---|
| Type | Embedding models |
| Task | Embeddings |
| Parameters (lead) | 209M |
| Context | 512 tokens |
| Runs with | sentence-transformers |
| Released | 2026-04-14 |
| Popularity | 1k downloads / month |
| Licence | Open weights |
About
Модель BERT для задач симметричного перефразирования (STS, поиск парафраз, дедупликация) и логического вывода (NLI). Получена дистилляцией эмбеддингов русских и английских текстов Qwen/Qwen3-Embedding-4B.
Модель может использоваться в качестве базовой для дообучения под пользовательские задачи классификации и кластеризации.
Основные характеристики модели:
from sentence_transformers import SentenceTransformer
model = SentenceTransformer('sergeyzh/rubert-large-uncased-sts')
sentences = ["привет мир", "hello world", "здравствуй вселенная"]
embeddings = model.encode(sentences)
print(model.similarity(embeddings, embeddings))
Оценки модели на задачах для русского языка:
| Model Name | RuSTS Benchmark STS | RU ParaPhraser STS | STS22,v2 | TERRa Classification | Average |
|---|---|---|---|---|---|
| Qwen3-Embedding-4B | 0,888 | 0,766 | 0,701 | 0,666 | 0,755 |
| rubert-large-uncased-sts | 0,869 | 0,771 | 0,686 | 0,664 | 0,748 |
| multilingual-e5-large-instruct | 0,840 | 0,754 | 0,706 | 0,639 | 0,735 |
| Qwen3-Embedding-0.6B | 0,842 | 0,721 | 0,662 | 0,607 | 0,708 |
| bge-m3 | 0,797 | 0,749 | 0,663 | 0,607 | 0,704 |
| multilingual-e5-base | 0,796 | 0,702 | 0,607 | 0,550 | 0,664 |
Оценки модели на задачах для английского языка:
| Model Name | STS12 | STS13 | STS14 | STS15 | STS17 | STS22,v2 | STS Benchmark | Average |
|---|---|---|---|---|---|---|---|---|
| Qwen3-Embedding-4B | 0,866 | 0,944 | 0,909 | 0,938 | 0,918 | 0,730 | 0,937 | 0,892 |
| Qwen3-Embedding-0.6B | 0,830 | 0,918 | 0,871 | 0,914 | 0,855 | 0,718 | 0,911 | 0,860 |
| rubert-large-uncased-sts | 0,818 | 0,901 | 0,864 | 0,901 | 0,906 | 0,660 | 0,896 | 0,849 |
| multilingual-e5-large-instruct | 0,825 | 0,881 | 0,848 | 0,910 | 0,860 | 0,690 | 0,884 | 0,842 |
| bge-m3 | 0,787 | 0,796 | 0,790 | 0,878 | 0,796 | 0,700 | 0,849 | 0,800 |
| multilingual-e5-base | 0,767 | 0,780 | 0,766 | 0,882 | 0,783 | 0,646 | 0,856 | 0,783 |
From the published model card. Full card on the HuggingFace links in the sidebar.
Using it via the API
Once AxForge deploys rubert-large-uncased-sts for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (rubert-large-uncased-sts below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/embeddings \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"rubert-large-uncased-sts","input":"text to embed"}'
Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.