Brief-details: Compact 1B parameter Llama model optimized for web deployment using MLC format, supporting chat and REST API functionality with q4f32_1 quantization.
Brief-details: A compact Stable Diffusion model fine-tuned on LAION Improved Aesthetics dataset, optimized for efficient text-to-image generation at 256x256 resolution
Brief Details: Japanese Sentence BERT model trained on JSNLI dataset, optimized for sentence similarity tasks with 523K training samples achieving 85.29% accuracy.
Brief-details: RoBERTa-based model fine-tuned for legal contract analysis and question answering, optimized for the CUAD dataset with improved AUPR score of 46.6%
BRIEF DETAILS: ELECTRA base generator model by Google - A self-supervised language model trained as a discriminator, optimized for efficient pre-training on limited compute resources.
Brief Details: DeepSeek MoE 16B Base - A powerful 16.4B parameter Mixture-of-Experts language model supporting commercial use, optimized for text generation tasks.
Brief Details: SmolLM-360M is a 362M parameter language model optimized for efficiency, trained on high-quality educational content with state-of-the-art performance for its size class
Brief Details: LLaMA-7B HuggingFace port - 7B parameter transformer-based LLM from Meta AI. Research-focused with strong reasoning capabilities across multiple languages.
Brief Details: Jamba-v0.1: A 51.6B parameter hybrid SSM-Transformer LLM with 256K context length, combining Mamba architecture with MoE for enhanced throughput.
Brief-details: A powerful CLIP vision-language model trained on LAION-2B dataset, achieving 78.4% ImageNet accuracy. Excels at zero-shot classification and image-text tasks.
Brief-details: RoBERTa-Large-based fact-checking model that determines if claims are supported by reference documents, achieving high accuracy on LLM-AggreFact benchmark.
BRIEF DETAILS: Qwen2.5-72B is a powerful 72.7B parameter language model with 128K token context length, supporting 29+ languages and optimized for diverse tasks including coding and mathematics.
Brief-details: E5-base is a 109M parameter text embedding model optimized for semantic similarity tasks, featuring 12 layers and 768d embeddings with MTEB benchmarks.
BRIEF-DETAILS: Facebook's English-to-German translation model using FSMT architecture. Achieves 42.83 BLEU score. Popular choice with 19.5k+ downloads.
BRIEF DETAILS: English Text-to-Speech model (36.3M params) from Facebook's MMS project using VITS architecture. Features non-deterministic speech generation with stochastic duration prediction.
Brief Details: LLaVA v1.6 Vicuna (13B params) - Advanced multimodal vision-language model for research, combining image understanding with text generation capabilities.
Brief Details: Yi-1.5-9B is a powerful 8.83B parameter language model trained on 3.6T tokens, offering strong performance in coding, math, and reasoning tasks.
Brief Details: PhotoMaker-V2 is an advanced text-to-image model specializing in customized photo generation from reference face images, built by TencentARC with SDXL compatibility.
Brief Details: Cutting-edge multimodal chatbot leveraging LLaMA/Vicuna with region-level visual understanding, trained on 1.7M+ data points
Brief-details: Korean RoBERTa-based sentence embedding model optimized for semantic similarity tasks, achieving 83.65% avg performance on STS benchmarks. 111M params.
Brief-details: A 3.09B parameter GGUF-formatted language model optimized for text generation with multiple quantization options (2-8 bit), based on Mistral architecture
