Brief-details: Swin-B is a hierarchical vision transformer using shifted windows, achieving strong performance on ImageNet with 83.5% accuracy and 88M parameters.
Brief-details: 72.7B parameter chat model optimized for conversation, available in multiple GGUF quantizations for efficient deployment across different hardware configurations
Brief Details: Russian-optimized 32B parameter LLM with custom tokenizer, offering up to 60% faster Russian text generation compared to base Qwen-2.5-32B-Instruct
BRIEF DETAILS: 32B parameter creative writing model focused on reducing repetition through unique dataset curation and unconventional single-epoch training approach using RS-LORA+.
Brief Details: Lightweight 1.54B parameter coding-focused model based on Qwen2.5, optimized for GGUF format with multiple quantization options (Q4/Q5/Q8) for efficient deployment.
Brief Details: A 12.2B parameter Mistral-based model fine-tuned to reduce archaic language while maintaining uncensored capabilities using ORPO tuning approach.
Brief Details: A 32.8B parameter code-specialized model built on Qwen2.5-Coder, featuring continuous finetuning and multiple quantization options for enhanced coding capabilities.
Brief-details: Qwen2.5-Coder-14B-Instruct-GGUF is a GGUF-quantized version of Qwen's 14B parameter code-specialized LLM, optimized for code generation and reasoning with 128K context window.
Brief Details: Quantized version of Qwen2.5-Coder-32B-Instruct using ExLlamaV2, offering multiple compression levels from 2.2 to 8.0 bits per weight for efficient deployment.
Brief Details: A fine-tuned ViT-GPT2 model (239M params) for image captioning, combining Vision Transformer and GPT-2 to generate natural language descriptions from images.
Brief Details: A 1.3B parameter SQL generation model optimized for text-to-SQL tasks, achieving competitive performance against larger models with efficient GGUF quantization.
Brief Details: Qwen2.5-Coder-32B-Instruct-8bit is a MLX-optimized coding model with 9.22B parameters, offering 8-bit precision for efficient code generation and chat capabilities.
Brief-details: Quantized version of Kaiju-11B optimized for GGUF format, featuring 10.7B parameters. Aimed at reducing GPT-like behaviors with merged architecture from multiple base models.
Brief-details: Multi-modal LLM based on Llama3 architecture (8.03B params) optimized for sound and text understanding, with strong performance on audio benchmarks
Brief-details: EndConvo-health-1b-GGUF-v1 is a 1.5B parameter LLM specialized in detecting conversation endpoints in healthcare dialogues, based on Llama-3.2-1B
BRIEF-DETAILS: Portuguese language model with 6.74B parameters, optimized for text generation and NLP tasks. GGUF quantized version of Sabiá-7B with strong performance on Portuguese benchmarks.
Brief-details: LLaMA-Pro-8B-Instruct-GGUF is a specialized 8.3B parameter model focused on programming and mathematical reasoning, with GGUF optimization for efficient deployment
BRIEF DETAILS: A specialized LoRA model trained on Flux.1-dev for generating pastel backgrounds, optimized for 1024x1024 resolution with 18 hi-res training images and constant LR scheduling.
BRIEF DETAILS: Stable Diffusion 3.5 Medium GGUF - An 8.27B parameter text-to-image model featuring MMDiT-X architecture with multi-resolution capabilities and enhanced prompt understanding.
Brief Details: A specialized LoRA model for FLUX.1 that transforms images into retro pixel art style, trained on 16 high-res images with 64 network dimensions and 15 epochs.
Brief Details: 8B parameter LLaMA-3 based uncensored model optimized for creative writing, storytelling, and roleplay with 128k context window and vivid prose capabilities
