BRIEF-DETAILS: Neural machine translation model for English-Spanish using bidirectional GRU with attention mechanism. Features teacher forcing and dynamic batching. Achieves 3.527 final loss.
BRIEF-DETAILS: 32B parameter Qwen model with multiple GGUF quantizations (8-bit to 2-bit) optimized for different hardware and RAM configurations, using imatrix techniques.
Brief-details: Specialized 14B parameter medical LLM trained on 20T tokens, featuring innovative architecture and superior medical domain performance across 20+ departments. Excellent for clinical applications.
BRIEF-DETAILS: Character generation template system for AI roleplaying, featuring detailed persona creation with consistent formatting and lore-accurate outputs
Brief-details: UI-TARS-72B-DPO is a cutting-edge GUI interaction model with superior perception and reasoning capabilities, achieving 90.3% accuracy in GUI tasks
BRIEF-DETAILS: Text-to-image AI model optimized for aesthetic and cinematic image generation in just 4 steps. Supports bfloat16, fp8, GGUF formats with Apache 2 license.
BRIEF DETAILS: Optimized GGUF quantized version of DeepSeek 7B model with multiple compression variants (2-6.4GB), featuring imatrix quantization for improved performance.
Brief-details: DeepSeek-R1-bf16 is a BF16 precision conversion of the original DeepSeek-R1 model, featuring 671B total parameters and 37B activated parameters, specializing in advanced reasoning and mathematical problem-solving.
BRIEF-DETAILS: 8B parameter Llama-based distilled model from DeepSeek-R1, optimized with Unsloth's dynamic 4-bit quantization for efficient reasoning and math capabilities
Brief-details: LatentSync is a ByteDance AI model for lip-sync video generation, featuring U-Net and SyncNet architectures with integrated Whisper support and face detection capabilities.
Brief Details: DeBERTa-v3-based classifier for web text quality assessment. Trained on 328K lines, achieves 0.81 micro-F1. Developed by TurkuNLP for high-quality data filtering.
BRIEF DETAILS: 1.5B parameter math-focused LLM from NVIDIA, fine-tuned on Qwen for mathematical reasoning with Chain-of-Thought capabilities
Brief-details: UNI2-h is an academic-focused AI model from MahmoodLab, available under CC-BY-NC-ND 4.0 license, requiring institutional verification for access.
Brief-details: A specialized LoRA model fine-tuned from Flux Fill Dev for object removal tasks. Enables seamless object deletion from images using masks. Non-commercial use only.
BRIEF-DETAILS: Comprehensive GGUF quantization collection of Microsoft's phi-4 model, offering 25+ variants optimized for different hardware and memory constraints, from 4.49GB to 58.64GB.
BRIEF-DETAILS: Sa2VA-8B is an 8B parameter multimodal AI model combining SAM2 & LLaVA capabilities for advanced image/video understanding with dense object segmentation
Brief-details: TabPFN-v2-reg is a transformer-based foundation model for small tabular regression tasks, requiring no task-specific training and published in Nature 2025.
BRIEF DETAILS: Dolphin 3.0 is an 8B parameter open-source LLM based on Llama 3.1, optimized for general-purpose tasks including coding, math, and function calling. Offers local deployment with customizable system prompts.
BRIEF-DETAILS: Lightweight 1B parameter Llama model optimized for 8-bit quantization, designed for instruction-following tasks using MLX framework
Brief Details: RuBERT - Russian BERT model with 180M parameters, trained on Wikipedia & news data. Features MLM & NSP capabilities, based on BERT-base.
BRIEF DETAILS: TCR-BERT: Specialized transformer model for T-cell receptor sequence analysis, focusing on masked amino acid modeling and antigen binding classification. Built for immunological research.
