BRIEF-DETAILS: MacBERT-based Chinese spelling correction model with 102M parameters. Achieves SOTA on SIGHAN2015 with 89.91% F1-score. Specialized for detecting and correcting Chinese text errors.
Brief-details: DeepSeek-Coder-V2-Instruct-0724 is a 236B parameter MoE code model with 21B active params, offering superior coding capabilities across 338 programming languages with 128K context length.
Brief-details: A powerful 13B parameter GGML-quantized language model based on Vicuna, fine-tuned on GPT-4 generated datasets with multiple quantization options for CPU/GPU inference.
Brief Details: AutoCoder - A 33.3B parameter code generation model surpassing GPT-4 Turbo accuracy (90.9%) with automatic package installation capability.
Brief-details: A ColBERT-based model supporting 8k context length, built on JinaBERT architecture. Optimized for passage retrieval with competitive performance to ColBERTv2.
Brief-details: DeepSeek Math 7B Instruct is a specialized mathematical reasoning LLM optimized for step-by-step problem solving, supporting both English and Chinese inputs with chain-of-thought prompting.
Brief Details: ControlNet checkpoint for depth-aware image generation with FLUX.1-dev model. Optimized for 1024x1024 resolution, non-commercial use.
Brief-details: Powerful AI model for image segmentation that can generate high-quality object masks from various input prompts, trained on 11M images and 1.1B masks
BRIEF DETAILS: A specialized anime-style image generation model offering clean anatomically-correct outputs with stunning backgrounds, available as LoRA (18MB) or full model merged with Anything V4.5.
Brief-details: Bilingual Korean-English vision-language model based on LLaMA 3.1 with 8.35B parameters. Handles both text and image tasks while maintaining strong language capabilities.
Brief-details: Meissonic is a non-autoregressive text-to-image synthesis model optimized for high-resolution image generation on consumer GPUs using masked generative transformers.
Brief-details: MusicGen-Medium is a 1.5B parameter text-to-music AI model from Meta that generates high-quality music from text descriptions using transformer architecture.
Brief Details: A powerful 3.6B parameter Japanese language model trained on 312.5B tokens, featuring advanced tokenization and strong text generation capabilities.
Brief-details: ACertainty is an anime-style Stable Diffusion model optimized for fine-tuning and dreambooth training, featuring balanced output and reduced laion-aesthetic bias.
Brief Details: A comprehensive collection of LoRA models focused on Blue Archive character generation, featuring 20+ character models with consistent preview demonstrations and ControlNet integration.
Brief Details: Scientific language model with 6.7B parameters, trained on 106B tokens of scientific text. Specialized in academic tasks and citations.
BRIEF-DETAILS: Large-scale text summarization model (460M params) optimized for long documents up to 16K tokens, fine-tuned on BookSum dataset with strong ROUGE scores
Brief Details: A fine-tuned Whisper large-v3 model optimized for Chinese ASR, achieving 24-65% improvement over baseline on major benchmarks like AISHELL and WenetSpeech.
Brief-details: A 34B parameter LLM fine-tuned on Yi-34B using diverse datasets and DPO training. Features multi-format prompting and reduced censorship capabilities.
Brief-details: Custom templates and settings collection for SillyTavern chat platform, supporting 9 different formats including Gemini, ChatML, and Mistral variants. Community-driven with 98 likes.
BRIEF-DETAILS: A specialized 7B parameter language model fine-tuned for mathematical reasoning, achieving strong performance in chain-of-thought tasks and formal mathematics
