BRIEF-DETAILS: ChilloutMix-NI is a specialized image generation model by swl-models, available on HuggingFace, focused on high-quality artistic outputs.
Brief Details: Word2Vec model trained on Google News (100B words), offering 300-dimensional vectors for 3M words/phrases. Key for NLP tasks & semantic analysis.
Brief Details: GGUF-converted Tencent HunyuanVideo-I2V model optimized for ComfyUI, enabling efficient image-to-video generation with quantized performance.
Brief-details: Fine-tuned Phi-4 multimodal model specialized for Korean speech-to-text tasks, trained on 35K samples with impressive ASR performance (3.80% CER on zeroth-test).
Brief Details: A 32B parameter LLM created by merging Qwen/QwQ-32B and Qwen2.5-32B to reduce catastrophic forgetting and enhance performance.
Brief-details: Sombrero-QwQ-32B-Elite10 - A 32B parameter LLM optimized for efficient memory usage and structured text generation, supporting 35+ languages and long context windows up to 256K tokens.
Brief-details: A 32B parameter LLM optimized for efficient memory usage and reasoning, specializing in coding, mathematical problem-solving, and logical deduction with 256K context window.
Brief-details: GGUF-based workflow model from Lightricks with multiple drag-and-drop functionality for i2v, t2v transformations and various quantization options
Brief-details: QwQ-32B is a powerful 32.5B parameter reasoning model from the Qwen series, featuring 131K context length and specialized architecture for enhanced problem-solving capabilities.
Brief-details: A comprehensive collection of GGUF quantizations of Meta-Llama-3.1-8B-SurviveV3, offering various compression levels from 16GB to 2.95GB with different quality-size tradeoffs.
BRIEF DETAILS: An uncensored 8B parameter variant of IBM's Granite-3.2 instruction model, modified using abliteration to remove content restrictions
BRIEF-DETAILS: Advanced multimodal retrieval model achieving SOTA performance on composed image retrieval tasks. Built on MegaPairs dataset with 26M+ triplets and fine-tuned on MMEB benchmark.
Brief-details: Novel text-to-image synthesis model focusing on scene-consistent camera control, accepted at CVPR 2025. Created by pandaphd team for realistic photo generation.
BRIEF DETAILS: RWKV7-3B-siglip2 is a vision-language model combining RWKV7 architecture with SigLIP2 encoder, achieving strong performance on VQA tasks (78.3% on VQAV2).
Brief-details: Grammar error correction model from Gramformer library. Specialized in detecting and fixing grammatical mistakes in English text using transformer architecture.
Brief-details: DeepSeek-R1-Distill-Qwen-1.5B-GGUF is a lightweight 1.5B parameter distilled version of DeepSeek-R1, offering efficient reasoning capabilities and GGUF format compatibility for local deployment.
Brief Details: A compact variation of Mistral architecture, designed for experimental and educational purposes with randomized weights, useful for testing ML pipelines.
Brief Details: A minimalist Falcon model implementation under 10MB, ideal for testing and educational purposes. Created by fxmarty for lightweight experimentation.
BRIEF-DETAILS: SFR-Embedding-Code-2B_R is a 2B parameter embedding model by Salesforce for multilingual code/text retrieval, achieving 67.4% NDCG@10 on CoIR benchmark
Brief-details: PaliGemma 3B is Google's pre-trained vision-language model requiring license acceptance on HuggingFace, designed for multimodal tasks with 3B parameters
BRIEF-DETAILS: GPT-SoVITS is a pretrained model collection for voice synthesis and conversion, integrating GPT and SoftVC technologies for enhanced audio generation.
