Brief-details: State-of-the-art English language embedding model with 335M parameters, achieving top performance on MTEB benchmarks. Optimized for retrieval and similarity tasks.
Brief Details: A DETR-based transformer model with 28.8M parameters for table detection in documents, trained on PubTables1M dataset with MIT license.
Brief Details: Indonesian RoBERTa-based POS tagger achieving 96.25% accuracy on IndoNLU dataset. 124M params, MIT licensed, optimized for Indonesian text.
Brief Details: DistilHuBERT - Efficient speech representation model with 23.5M params. 75% smaller than HuBERT while maintaining performance. Ideal for academic/small-scale ML.
Brief Details: SigLIP vision-language model with 878M parameters optimized for zero-shot classification. Uses sigmoid loss, trained on WebLI dataset at 384x384 resolution.
Brief Details: BART-Large-MNLI is a 407M parameter NLI model fine-tuned for zero-shot classification, offering powerful multi-label text classification capabilities.
Brief-details: BGE-M3 is a versatile multilingual embedding model supporting dense retrieval, lexical matching, and multi-vector interaction across 100+ languages with 8192 token context.
BRIEF DETAILS: A compact yet powerful ColBERT-based retrieval model with 33.4M parameters, outperforming larger models in passage retrieval tasks while maintaining efficiency.
Brief Details: Audio Spectrogram Transformer with 86.6M params, fine-tuned on AudioSet. Converts audio to spectrograms for classification using ViT architecture.
Brief Details: BART base model (139M params) by Facebook - A transformer-based seq2seq model for text generation and comprehension tasks, pre-trained on English text.
Brief Details: BERTimbau Base - A state-of-the-art BERT model for Brazilian Portuguese with 110M parameters, trained on brWaC dataset for NLP tasks.
Brief-details: State-of-the-art multilingual speech recognition model with 1.54B parameters, supporting 99 languages and offering improved accuracy over previous versions.
Brief-details: Multilingual XLM-RoBERTa model (560M params) fine-tuned for token classification/NER, supporting 94 languages with strong performance on English CoNLL-2003 dataset.
Brief Details: BERT multilingual base model supporting 102 languages, 168M parameters, uncased tokenization, ideal for masked language modeling and sequence tasks.
Brief Details: A fine-tuned XLSR-53 model for Russian speech recognition, achieving 13.3% WER on Common Voice, with 3.6M+ downloads and Apache 2.0 license.
Brief-details: Vision Transformer model with 86.6M params for image classification, pre-trained on ImageNet-21k and fine-tuned on ImageNet-1k. Popular with 3.7M+ downloads.
Brief-details: A powerful multilingual speech model with 300M parameters, supporting 126 languages. Pre-trained on 436K hours of audio data using wav2vec 2.0 architecture.
BRIEF DETAILS: Clinical BERT model trained on MIMIC III healthcare data, combining BioBERT initialization with clinical note training for specialized medical NLP tasks.
Brief-details: A fine-tuned Vision Transformer (ViT) model with 85.8M parameters for NSFW image classification, achieving 98% accuracy using an 80k image dataset.
Brief-details: A transformer-based semantic segmentation model fine-tuned for clothes and human parsing, offering 27.4M parameters with strong accuracy (80% mean) across 18 clothing categories.
Brief-details: A powerful multilingual language detection model supporting 20 languages with 99.6% accuracy, based on XLM-RoBERTa, featuring 278M parameters.
