Brief Details: SummLlama3-8B: An 8B parameter summarization model outperforming Llama3-70B and GPT-4, specialized in human-preferred summaries across 7 domains.
Brief-details: CompassJudger-1-32B-Instruct is a comprehensive evaluation model based on Qwen2.5-32B, specialized in scoring, comparing, and reviewing AI model outputs with formatted assessment capabilities.
Brief-details: A UI-focused multimodal LLM (8.4B params) built on Llama-3-8B for UI tasks, featuring image-text capabilities and specialized grounding functionality.
Brief-details: FluxFit is an innovative virtual fitting AI model based on the Flux architecture, offering non-commercial virtual try-on capabilities with support for ONNX and Safetensors.
Brief-details: PDF-Extract-Kit-1.0 is an open-source toolkit for PDF data extraction, offering streamlined integration via HuggingFace or Git with Apache 2.0 licensing
Brief Details: [756M parameter Japanese speech recognition model based on Whisper, 6.3x faster than large-v3 with competitive accuracy on Japanese ASR tasks]
Brief-details: MN-Halide-12b is a sophisticated 12.2B parameter language model created through Model Stock merging of 18 high-quality Mistral-based models
Brief-details: Advanced lip-syncing model combining Wav2Lip and Real-ESRGAN algorithms for high-fidelity video outputs with improved visual quality and accurate audio synchronization
Brief-details: Specialized Uzbek-English language model based on Mistral, optimized for translation, summarization & QA tasks with strong BLEU/COMET scores and maintained English capabilities.
Brief-details: Open-source Sora-like video generation model featuring WFVAE, prompt refining, and sparse attention. Supports 93x480p video generation within 24GB VRAM.
Brief-details: Text-to-image diffusion model trained on CC-licensed images, supports Japanese/English, based on PixArt-Σ architecture, requires 8GB+ VRAM GPU, Apache 2.0 licensed.
Brief-details: 72B parameter multilingual LLM fine-tuned on Claude-style data, supporting 9 languages with strong performance on IFEval (75.6%) and BBH (57.85%).
Brief-details: NovelAI's original anime-focused Stable Diffusion model, optimized for anime-style image generation with CLIP skip 2 and specialized UNet/VAE components.
Brief Details: A creative LoRA model for Stable Diffusion that generates vibrant illustrations with bold linework and retro-modern aesthetics, using FLUX.1-dev as base.
Brief Details: Creative writing-focused 8B parameter LLaMA-based model with low censorship, optimized for storytelling and instruction following. Available in multiple quantizations.
Brief-details: Revolutionary theorem-proving LLM achieving SOTA results (63.5% on miniF2F) using reinforcement learning and Monte-Carlo tree search for mathematical proofs.
Brief Details: Tanuki-8B-dpo is a bilingual (Japanese/English) 7.51B parameter LLM trained on 1.3T tokens, optimized for dialogue through DPO fine-tuning.
Brief Details: A 7.7B parameter fact-checking model that evaluates if claims are supported by documents, achieving SOTA performance with efficient throughput
BRIEF DETAILS: An 8B parameter LLM based on Meta-Llama-3.1, optimized for long-form content generation up to 10,000+ words with support for English and Chinese.
Brief Details: A specialized LoRA model for FLUX.1-dev that creates paper cutout-style artistic images, featuring high downloads and community recognition
Brief-details: Hermes-3 70B FP8 is a highly capable LLM using ChatML format, featuring function calling, JSON mode & advanced reasoning. Built on Llama 3.1.
