Models

Brief Details: Stable Video Diffusion XT - Advanced AI model that converts still images to 25-frame videos at 576x1024 resolution. Built by Stability AI with commercial licensing available.

Brief-details: Compact pre-trained model for multivariate time-series forecasting with <1M parameters. Outperforms larger models in zero-shot/few-shot forecasting for minutely/hourly data.

Brief-details: Text embedding model fine-tuned on MEDI dataset and MTEB Classification data, offering strong performance for semantic search and text similarity tasks without requiring instructions.

Brief Details: Stable Diffusion v2 inpainting model for image editing and generation. Built on SD2-base, enables selective image modification using masks. 557K+ downloads.

Brief-details: GPT-2 Medium (355M params) - Advanced language model by OpenAI for text generation, trained on web content with broad capabilities for AI research and creative applications.

Brief-details: Spanish sentiment analysis model based on BETO, trained on TASS 2020 tweets. Handles POS/NEG/NEU classifications with 565k+ downloads.

Brief Details: OpenSora-VAE-v1.2 is a 393M parameter video autoencoder model for high-quality video processing, part of the Open-Sora ecosystem

Brief-details: ViT-based image classification model achieving 99.23% accuracy, fine-tuned from google/vit-base-patch16-224-in21k with 85.8M parameters, Apache 2.0 licensed.

Brief Details: FLAN-T5 Base (248M params) - Instruction-tuned language model excelling at text generation, translation & reasoning tasks, supported across 5 languages.

Brief Details: Large-scale English Named Entity Recognition model built with Flair, achieving 94.36% F1-score on CoNLL-03. Specializes in 4-class entity detection.

Brief-details: Lightweight 3.8B parameter instruction-tuned LLM with strong multilingual capabilities, 128K context support, and competitive performance against larger models

BRIEF DETAILS: Table Transformer model for document table structure recognition. 28.8M params, MIT license, based on DETR architecture. High adoption with 577K+ downloads.

BRIEF DETAILS: Multilingual ALBERT model pre-trained on 12 Indian languages with 9B tokens, achieving state-of-the-art performance on various NLP tasks

BRIEF-DETAILS: Audio transformer model for synthetic voice detection with 86.2M parameters. Achieves 99.99% accuracy in detecting AI-generated voices. Based on MIT's AST architecture.

Brief Details: A sentence embedding model that maps text to 768-dimensional vectors, based on DistilBERT architecture with 66.4M parameters. Optimized for semantic search and clustering.

Brief Details: Qwen2.5-0.5B-Instruct: 494M parameter instruction-tuned LLM with 32K context window, optimized for multilingual tasks and structured outputs

Brief Details: Multilingual translation model supporting 50 languages with 611M parameters. Facebook's mBART-50 variant for direct translation between language pairs.

Brief-details: Vision Transformer model fine-tuned for facial emotion recognition, achieving 71.16% test accuracy. Supports 7 emotions using ViT architecture with 85.8M parameters.

Brief-details: High-performance German-to-English translation model from Helsinki-NLP, achieving BLEU scores of up to 43.7 on newstest2018, built on transformer architecture

Brief-details: A French-to-English translation model by Helsinki-NLP using the Marian framework, achieving BLEU scores of 30-57 across various test sets with robust performance on news content.

Brief-details: A Helsinki-NLP Arabic-to-English translation model built on the OPUS dataset using transformer architecture, achieving 49.4 BLEU score on Tatoeba benchmark.