Brief-details: Skywork-13B-base is a powerful bilingual LLM with 13B parameters, trained on 3.2T tokens of Chinese/English data, achieving SOTA performance in various benchmarks.
Brief-details: Qwen-7B-Chat-Int4 is a 4-bit quantized version of the Qwen-7B-Chat model, offering efficient inference with 2.11B parameters while maintaining strong performance across multiple languages and tasks.
BRIEF-DETAILS: 2.7B parameter chat model specialized in function calling, built on replit-code-v1-3b, with GPT-3.5/4-like capabilities for API integration
Brief-details: A 7B parameter uncensored Llama2 chat model quantized to 4-bit precision, offering multiple GPTQ variants for efficient GPU inference and deployment
Brief Details: A 7B parameter language model fine-tuned by Open-Assistant on human demonstrations, based on StableLM, optimized for assistant-style conversations.
Brief Details: Core ML implementation of Stable Diffusion XL base model, optimized for macOS GPUs with ORIGINAL attention implementation. Created by Apple, supporting text-to-image generation.
Brief-details: An artistic AI model focusing on balanced scene and character generation, featuring strong light/shadow control and improved tag reading in v5.0. Supports both landscape and portrait modes.
Brief-details: Japanese art-focused diffusion model built on Defacta base, optimized for anime/manga-style images with multiple versions (V1-V9) and specialized VAE requirements
Brief Details: StreetCLIP - A powerful zero-shot image geolocalization model trained on 1.1M street-level images, achieving SOTA performance in geographic classification tasks.
Brief-details: A comprehensive collection of mixed anime/weeb Stable Diffusion models, featuring 9 different model combinations optimized for high-quality anime-style image generation.
Brief-details: A compact multilingual translation model supporting 101 languages with 333M parameters, achieving M2M-100 comparable performance while being 3.6x smaller.
Brief-details: Moebius-style art concept for Stable Diffusion, MIT-licensed textural inversion model enabling Jean Giraud's distinctive sci-fi art style generation.
Brief-details: CyberAgent's 22B parameter bilingual (Japanese/English) chat model with 16K context window, trained on 2T tokens and fine-tuned for dialogue use cases.
Brief-details: A powerful 7B parameter LLM with 1M token context window, excelling in math reasoning and long-text comprehension. Features state-of-the-art performance and advanced tool utilization capabilities.
Brief Details: CogVLM2 is a powerful 19.5B parameter vision-language model supporting 8K text length and 1344x1344 image resolution with Chinese/English capabilities
Brief Details: TURNA - A 1.14B parameter Turkish language model based on UL2 framework, optimized for text generation and understanding tasks with 36 layers and 16 attention heads.
BRIEF DETAILS: 72B parameter LLM using DPO training on QWEN base model. Features LoRA optimization, MIT license, AMD MI250 GPU compatible. Focused on text generation tasks.
Brief Details: Crystal is a 7B parameter LLM trained on SlimPajama and StarCoder, excelling in both natural language and coding tasks with competitive benchmark performance.
Brief-details: A powerful 33B parameter code generation model trained on 2T tokens (87% code, 13% language), supporting multiple programming languages with 16K context window.
BRIEF DETAILS: 7B parameter LLaMA-2 based model fine-tuned on 300k+ instructions, optimized for long responses and reduced hallucination, MIT licensed
Brief Details: Korean-optimized LLaMA-2 chat model (7B params) fine-tuned on KULLM-v2 dataset, offering improved Korean language capabilities
