Model reference · open weights

BERTA

Available as managed deployment Embeddings sergeyzh · community Embeddings 1 variants 18k dl/mo

BERTA 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 bysergeyzh
TypeEmbedding models
TaskEmbeddings
Parameters (lead)128M
Context512 tokens
Runs withsentence-transformers
Based onsergeyzh/LaBSE-ru-turbo
Released2025-03-10
Popularity18k downloads / month
LicenceOpen weights

About

What BERTA is

BERTA

Модель для расчетов эмбеддингов предложений на русском и английском языках получена методом дистилляции эмбеддингов ai-forever/FRIDA (размер эмбеддингов - 1536, слоёв - 24) в sergeyzh/LaBSE-ru-turbo (размер эмбеддингов - 768, слоёв - 12). Основной режим использования FRIDA - CLS pooling заменен на mean pooling. Каких-либо других изменений поведения модели не производилось. Дистиляция выполнена в максимально возможном объеме - эмбеддинги русских и английских предложений, работа префиксов.

Размер контекста модели соответствует FRIDA - 512 токенов.

Read the full model card

Префиксы

Все префиксы унаследованы от FRIDA. Оптимальный (обеспечивающий средние результаты) префикс для большинства задач - "categorize_entailment: " прописан по умолчанию в config_sentence_transformers.json

Перечень используемых префиксов и их влияние на оценки модели в encodechka:

ПрефиксSTSPINLISATI
-0,8420,7570,4630,8300,985
search_query:0,8530,7670,4790,8250,987
search_document:0,8310,7490,4630,8170,986
paraphrase:0,8470,7780,4460,8250,986
categorize:0,8570,7650,5010,8290,988
categorize_sentiment:0,5890,5350,4170,8050,982
categorize_topic:0,7400,5210,3960,7700,982
categorize_entailment:0,8410,7620,5710,8270,986

Задачи:

  • Semantic text similarity (STS);
  • Paraphrase identification (PI);
  • Natural language inference (NLI);
  • Sentiment analysis (SA);
  • Toxicity identification (TI).

Метрики

Оценки модели на бенчмарке ruMTEB:

Model NameMetricFRIDABERTArubert-mini-fridamultilingual-e5-large-instructmultilingual-e5-large
CEDRClassificationAccuracy0.6460.6220.5520.5000.448
GeoreviewClassificationAccuracy0.5770.5480.4640.5590.497
GeoreviewClusteringP2PV-measure0.7830.7380.6980.7430.605
HeadlineClassificationAccuracy0.8900.8910.8800.8620.758
InappropriatenessClassificationAccuracy0.7830.7480.6980.6550.616
KinopoiskClassificationAccuracy0.7050.6780.5950.6610.566
RiaNewsRetrievalNDCG@100.8680.8160.7210.8240.807
RuBQRerankingMAP@100.7710.7520.7110.7170.756
RuBQRetrievalNDCG@100.7240.7100.6540.6920.741
RuReviewsClassificationAccuracy0.7510.7230.6580.6860.653
RuSTSBenchmarkSTSPearson correlation0.8140.8220.8030.8400.831
RuSciBenchGRNTIClassificationAccuracy0.6990.6900.6250.6510.582
RuSciBenchGRNTIClusteringP2PV-measure0.6700.6500.5860.6220.520
RuSciBenchOECDClassificationAccuracy0.5460.5550.4930.5020.445
RuSciBenchOECDClusteringP2PV-measure0.5660.5560.5070.5280.450
SensitiveTopicsClassificationAccuracy0.3980.3990.3730.3230.257
TERRaClassificationAverage Precision0.6650.6570.6060.6390.584
Model NameMetricFRIDABERTArubert-mini-fridamultilingual-e5-large-instructmultilingual-e5-large
ClassificationAccuracy0.7070.6980.6310.6540.588
ClusteringV-measure0.6730.6480.5970.6310.525
MultiLabelClassificationAccuracy0.5220.5100.4630.4120.353
PairClassificationAverage Precision0.6650.657

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys berta for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (berta 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":"berta","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.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms