BRIEF-DETAILS: DeepSeek-R1-Distill-Qwen-7B-6bit is a 6-bit quantized MLX format model derived from DeepSeek-R1-Distill-Qwen-7B, optimized for efficient deployment
BRIEF DETAILS: A specialized Stable Diffusion image generation model focused on specific anatomical views, utilizing trigger words for consistent results.
BRIEF DETAILS: 24B parameter Mistral-based model optimized for reasoning, with multiple GGUF quantization options (6.5GB-94GB). Features special prompt format & reasoning system prompt.
BRIEF DETAILS: Sentence embedding model that converts text to 384-dim vectors, optimized for semantic search & clustering. Based on MiniLM architecture with strong performance on MS MARCO dataset.
Brief Details: Efficient sentence embedding model that maps text to 384-dimensional vectors. Fast, lightweight implementation of MiniLM architecture optimized for semantic search and clustering.
Brief-details: Russian-specific sentence encoder that maps text to 768D vectors. Built on DeBERTa-v1-base with 85M params. Optimized for semantic search & clustering.
Brief Details: Compact version of Gemini model extracted from Chrome browser, loadable via MediaPipe framework. Features basic instruction tuning capabilities.
brief-details: German-optimized DistilBERT model with cased tokenization, offering efficient NLP capabilities for German language tasks through knowledge distillation
Brief-details: CTRL - Salesforce's 140GB conditional transformer language model with controllable text generation via domain-specific codes. Supports creative writing and NLP research.
Brief-details: BERT large cased model with whole word masking, 336M parameters, 24 layers, trained on BookCorpus and Wikipedia. Optimized for bidirectional language understanding.
Brief-details: German BERT model (uncased) trained by DBMDZ team, optimized for German language processing tasks with transformer architecture and full vocabulary support
Brief Details: An 8B parameter AI model focused on card-based projections and analysis, developed by AlexBefest. Available on HuggingFace for preview access.
BRIEF-DETAILS: Lightweight 70M parameter chat model trained on Discord data, featuring Llama-3 architecture and customizable chat styles through usernames. Optimized for casual conversations.
Brief-details: A Llama-3.1-8B variant fine-tuned with DPO for enhanced emoji expression and friendly interactions. 8B parameters, optimized for engaging responses.
Brief-details: A 32B parameter merged LLM combining QwQ, DeepSeek-R1, and Light-R1 models, achieving superior performance on math reasoning and AIME benchmarks
Brief-details: Spanish language instruction-tuned 1B parameter LLaMA model fine-tuned on UNAL academic Q&A dataset using LoRA adaptation, optimized for academic text generation
Brief-details: Enhanced 35B parameter model combining QwQ-32B with TinyR1 and DeepSeek capabilities. Optimized for reasoning, creative generation, and instruction following. Requires ChatML template.
Brief-details: Glowing-Forest-12B-i1-GGUF is a quantized version of the original Glowing-Forest-12B model, offering various compression options from 3.1GB to 10.2GB with different quality-performance tradeoffs.
Brief-details: MN-Sappho-n3-12B-GGUF is a quantized version of the MN-Sappho model offering various compression levels, with sizes ranging from 4.9GB to 13.1GB, optimized for different performance/quality trade-offs.
Brief-details: A specialized LoRA model focused on CFG distillation for Stable Diffusion, created by spacepxl for enhanced image generation control and quality optimization.
BRIEF DETAILS: 12B parameter quantized language model with multiple GGUF variants (3.1GB-10.2GB), optimized for different performance/quality tradeoffs using imatrix quantization.
