Brief-details: A specialized anime-style Flux Dev model for text-to-image generation, featuring a Dim/Rank 64 architecture and unique "sftsrv style" trigger word. Popular with 8,027 downloads.
Brief-details: Qwen2's 57B MoE model with 14B active parameters. Excels in language tasks with 65k context length. Strong performance in coding, math & multilingual tasks.
Brief Details: Llama-3 based 8B parameter model specialized in conversational AI with balanced RP/ERP capabilities, trained on diverse datasets including Aesir and NoRobots.
Brief Details: Saul-7B-Instruct-v1: A specialized 7B parameter legal domain LLM built on Mistral-7B, offering robust legal text generation capabilities with MIT license.
Brief Details: Enhanced version of ChatGLM3-6B with 128K context length support. Optimized for long-text processing with improved position encoding and specialized training.
Brief Details: A 13B parameter LLaMA2-based merged model optimized for creative writing and fiction, with medical knowledge integration. Features multi-format instruction support.
BRIEF DETAILS: Self-RAG 7B model that combines text generation with self-reflection capabilities, built on LLaMA 2 architecture with MIT license
Brief-details: WhisperKit Core ML implementation for efficient speech recognition, featuring optimized variants with different file sizes and WER trade-offs. Achieves 2.44% WER on LibriSpeech.
Brief-details: OpenHermes-2-Mistral-7B-GGUF is a high-performance 7B parameter GGUF model optimized for chat and instruction-following, trained on 900k GPT-4 generated entries
Brief-details: A powerful 13B parameter language model trained on 2.6T tokens, supporting both English and Chinese, achieving SOTA performance across multiple benchmarks.
BRIEF-DETAILS: Baichuan2-7B-Base is a powerful open-source LLM trained on 2.6T tokens, supporting both Chinese and English, achieving state-of-the-art performance in its size class.
Brief-details: A 3B parameter multilingual LLM fine-tuned on xP3 dataset, capable of instruction-following across 46 languages with strong zero-shot performance.
BRIEF DETAILS: A 6B parameter decoder-only Transformer model specialized in code generation, supporting multiple languages with infilling capabilities. Created by Facebook.
Brief Details: A specialized Stable Diffusion model fine-tuned for creating balloon art images, featuring text-to-image capabilities and optimized for twisted balloon designs.
BRIEF-DETAILS: Specialized Stable Diffusion model fine-tuned for generating microscopic imagery. Features text-to-image capabilities with optimization for scientific visualization.
Brief-details: Koala is an academic dialogue model built on LLaMA, designed for research purposes with training data from OpenAI and ShareGPT. Features strong conversational abilities with academic focus.
Brief Details: Russian GPT-3 large language model trained on 80B tokens, achieving 13.6 perplexity. Specialized for Russian text generation with 2048 context length.
Brief Details: Compact multilingual transformer model supporting 16 languages, optimized through distillation with 21M parameters for efficient cross-lingual NLP tasks.
BRIEF-DETAILS: Advanced image captioning model based on Florence-2, with 271M params. Features detailed caption generation and tag creation for AI art workflows.
Brief Details: A massive 1T parameter multilingual LLM built on decoder-only architecture, trained on 2.3T tokens with enhanced factual capabilities and efficient pre-training
Brief Details: RYS-XLarge: 78B parameter AI model utilizing innovative layer duplication technique achieving 79.96% IFEval accuracy. MIT licensed.
