Brief Details: Cutting-edge SDXL-based generative model supporting 1536×1536 native resolution with hybrid NLP/tag-based prompting. Designed for high-fidelity illustration generation.
BRIEF DETAILS: Powerful 21B parameter MOE model combining 8 LLaMA 3.2 experts. Uncensored, creative writing focused with 128k context. Features Brainstorm 5x enhancement and exceptional instruction following.
Brief-details: Lightweight 500M parameter multimodal model for video/image analysis. Efficient (1.8GB GPU RAM), supports video QA, captioning, and comparison tasks.
Brief-details: DeepNovel-7B-Roleplay: A 7B parameter LLM fine-tuned from Qwen-7B-Instruct, optimized for roleplay, character development, and novel-length text generation with enhanced context retention.
Brief-details: An 8B parameter LLM from IBM with enhanced reasoning, multi-language support, and long-context capabilities. Features controllable thinking and strong performance on reasoning tasks.
Brief Details: AIFS Single v1.0 is ECMWF's operational AI weather forecasting system using GNN and transformer architecture for accurate global weather predictions
BRIEF-DETAILS: RuModernBERT-base: 150M parameter Russian BERT model trained on 2T tokens with 8K context length, optimized for masked language modeling
BRIEF-DETAILS: Large-scale RL-trained reasoning model (671B params) without supervised fine-tuning, specialized in math, code & reasoning tasks. Remarkable chain-of-thought capabilities.
Brief-details: Smallest multimodal vision-language model (256M params) capable of image understanding and text generation. Runs on 1GB GPU RAM, optimized for efficiency.
BRIEF-DETAILS: A model by Isam03 hosted on HuggingFace, currently with limited public information available. Further details and specifications pending.
Brief-details: Vision Transformer model for human pose estimation achieving 81.1 AP on COCO. Scales from 100M-1B parameters with simple architecture.
Brief-details: Huginn-0125 is a 3.5B parameter latent recurrent-depth model trained on 800B tokens, featuring variable computation depth and adaptive per-token compute capabilities.
Brief-details: RWKV-7 World is a multilingual language model trained on 3.1T tokens across 100+ languages, featuring improved MMLU scores and flexible architecture sizes from 0.1B to 7B parameters.
Brief-details: CNPM is an advanced AI model for recognizing Chinese pentatonic modes (Gong, Shang, Jiao, Zhi, Yu) in traditional music using spectral analysis and deep learning techniques.
Brief-details: An AI model for recognizing 8 distinct Guzheng (traditional Chinese instrument) playing techniques, trained on 2,824 audio clips with 85.5% accuracy using ViT architecture
Brief-details: CTIS is a specialized AI model for identifying Chinese traditional instruments, covering 200+ instrument types with high accuracy across different sound features (Mel: 97.3%, CQT: 98%).
BRIEF DETAILS: GemmaX2-28-9B-v0.1 is a 9B parameter multilingual translation model supporting 28 languages, built on Gemma2-9B with extensive pretraining on 56B tokens.
Brief-details: SVDQuant-based 4-bit quantized FLUX.1-dev model achieving 3.6× memory reduction and 8.7× speedup, optimized for efficient image generation on consumer GPUs
Brief Details: Advanced multilingual TTS model supporting 13 languages with extensive training data (1M+ hours). Best coverage for English/Chinese (300k+ hours each). Research-backed with BY-CC-NC-SA-4.0 license.
Brief Details: Timer-base-84m is an 84M-parameter generative Transformer for time series forecasting, pre-trained on 260B time points with zero-shot capabilities
Brief-details: claude-monet is a FLUX-based LoRA model specialized in generating Monet-style water lily paintings, trained to replicate the artist's iconic impressionist style.
