Brief-details: A custom AST (Abstract Syntax Tree) testing model by Ericwang, designed for experimental validation of Hugging Face repository functionality.
Brief-details: Kokoro-82M is a compact yet powerful text-to-speech model with 82M parameters, supporting multiple voices and quantization options for efficient deployment.
Brief-details: A lightweight variant of Qwen2VL model with randomly initialized weights, designed for experimental purposes in vision-language tasks
BRIEF DETAILS: Experimental AI motion stabilization model with two variants (mid/high), offering different stability-movement tradeoffs for animation applications
Brief-details: DeepSeek-V3-AWQ is a quantized version of DeepSeek V3, optimized for efficient inference on GPU clusters with AWQ, offering 48 TPS on 8x H100 and 38 TPS on 8x A100 GPUs.
Brief Details: A 70B parameter Japanese-focused multilingual model trained on 40M news articles. Features 128k context window, cross-lingual capabilities (JP/KR/CN/EN), and enhanced logical reasoning.
BRIEF-DETAILS: YarnGPT is a specialized text-to-speech model generating Nigerian-accented English, featuring 11 distinct voices and built on HuggingFaceTB/SmolLM2-360M architecture.
BRIEF-DETAILS: 3B parameter Llama-based model optimized for general-purpose tasks, coding, and math. Features ChatML format and customizable system prompts.
Brief-details: A re-implementation of the pi0 vision-language-action model from Physical Intelligence (Pi), focused on bridging visual perception and action planning.
Brief Details: A 12B parameter GGUF quantized model optimized for ARM architecture, featuring Imatrix compression for efficient deployment and inference.
BRIEF-DETAILS: 1B parameter Llama-based assistant model focused on general-purpose tasks, coding, and math. Features ChatML format and customizable system prompts.
Brief-details: Dolphin3.0-Qwen2.5-0.5B is a 500M parameter general-purpose AI assistant model focused on coding, math, and function calling, built on Qwen architecture.
BRIEF DETAILS: A powerful 32B parameter multilingual model fine-tuned from Qwen2.5, specialized in Japanese/Korean/Chinese/English tasks with enhanced logical reasoning and math capabilities.
Brief-details: A specialized 3B parameter medical verification model designed to evaluate the correctness of LLM outputs on medical problems, developed by FreedomIntelligence.
Brief-details: Diogenes-12B is a 12 billion parameter language model based on Mistral v3 formatting, currently in experimental phase and untested. Developed by Nitral-Archive.
BRIEF-DETAILS: Advanced 8B medical LLM built on LLaMA-3.1, specializing in complex medical reasoning with think-before-answer approach and English language support
Brief-details: Advanced 24B parameter Mistral MOE model combining 4x7B models, optimized for creative writing and roleplay. Features uncensored output, high-precision float32 architecture, and 32k context window.
Brief-details: Codepy-Deepthink-3B-GGUF is a fine-tuned LLaMA 3.2 variant optimized for deep reasoning and programming tasks, offering multiple quantization options and integration with popular inference platforms.
Brief-details: TinySwallow-1.5B is a compact 1.5B-parameter Japanese language model created through TAID distillation from Qwen2.5-32B, featuring enhanced Japanese capabilities
Brief Details: Finetuned version of LTX Video 0.9 VAE focused on reducing checkerboard artifacts through decoder and encoder improvements while maintaining latent space compatibility
Brief Details: Mia-LLM: A Mistral-7B-based personal AI assistant offering translation in 96 languages, document processing, and interactive services with advanced analysis capabilities.
