Brief Details: Norwegian BERT-based NER model (124M params) trained on NorNE dataset, specialized for named entity recognition in Norwegian text
Brief Details: A German language semantic similarity model with 336M parameters, achieving SOTA performance on STS benchmarks with 86.26% Spearman correlation.
Brief Details: 4.98B parameter INT4 quantized visual question-answering model, optimized for efficient GPU usage (9GB) with multimodal capabilities.
Brief-details: A 406M parameter BART-based model fine-tuned for meeting summarization, achieving 53.8% ROUGE-1 score on validation sets. Optimized for dialogue and meeting context.
Brief Details: ProLong 8B parameter LLM with 512K token context window, fine-tuned from Llama-3. Optimized for long-context tasks with extensive training on specialized datasets.
Brief-details: Qwen2.5-72B-Instruct GGUF - A powerful 72B parameter language model with multiple quantization options, optimized for different hardware configurations and use cases
Brief-details: RoBERTa-based model trained on 154M tweets for targeted sentiment analysis, offering 5-class classification from strongly negative to strongly positive
Brief-details: Snowflake's 33.2M parameter text embedding model optimized for retrieval tasks, achieving SOTA performance with 384-dim embeddings and NDCG@10 of 51.98
Brief Details: A powerful multilingual translation model supporting 16 Romance languages to English, with strong BLEU scores (54.9 average) and extensive testing
Brief-details: OWL-ViT is a zero-shot text-conditioned object detection model using CLIP backbone with ViT-L/14 architecture, enabling open-vocabulary object detection through text queries.
Brief-details: BERT-based question-answering model trained on SQuAD v2, achieving 71.15% exact match accuracy. Optimized for extractive QA tasks in English.
Brief-details: A lightweight 4-layer toxicity classifier (38.5M params) by IBM for detecting harmful content, optimized for low latency and CPU deployment with RoBERTa architecture.
Brief Details: A powerful 32.5B parameter LLM with 128K context, supporting 29+ languages. Excels in coding, math & long-form content generation.
Brief Details: Fast NER model for English using Flair, achieving 89.3% F1-score on Ontonotes dataset. Identifies 18 entity types with LSTM-CRF architecture.
Brief Details: LLaVA-OneVision 72B parameter multimodal model supporting English/Chinese, with strong performance on visual tasks (93.5% DocVQA accuracy). Built on Qwen2 architecture.
Brief-details: 8B parameter Llama-based roleplay model optimized for creative writing and diverse character interactions, featuring GGUF quantization and ARM compatibility
BRIEF DETAILS: A unified ControlNet model for FLUX.1-dev supporting 7 control modes with enhanced training. Optimized for high-quality image generation with multiple control types.
BRIEF-DETAILS: Chinese BERT model featuring Whole Word Masking, optimized for Chinese NLP tasks. Created by HFL team with 24K+ downloads, Apache 2.0 licensed.
Brief-details: ONNX-optimized RoBERTa model for emotion detection with 28 emotion categories. Features both full precision and INT8 quantized versions for faster inference and smaller size.
Brief Details: OmniGenXL is a versatile text-to-image model supporting both NSFW/SFW content generation, featuring ultra-realistic output with high-quality rendering capabilities at 512x512 resolution.
Brief Details: RoBERTa-based model fine-tuned for multi-label tweet topic classification, achieving 0.76 F1 score. Built by CardiffNLP with 24K+ downloads.
