Brief-details: A 7.62B parameter merged LLM combining Qwen2.5 variants optimized for creative writing, technical tasks & math reasoning. Achieves 77.08% on IFEval & strong instruction-following capabilities.
Brief-details: 32.5B parameter uncensored language model with various GGUF quantization options (11-32GB), trained on 7 datasets including WizardLM and SuperCOT
Brief Details: Optimized 17.1B parameter AWQ-quantized version of Mistral's latest LLM, offering multilingual support, coding expertise, and advanced reasoning capabilities.
Brief-details: A 12.2B parameter TIES-merged model combining Chronos-Gold-12B and ChatWaifu_Magnum for enhanced conversational capabilities, utilizing FP16 precision.
Brief Details: An 8B parameter merged LLM combining Aspire, Heart Stolen, and CursedMatrix models, optimized for creative writing and general tasks. BF16 precision.
Brief-details: Japanese Text-to-Speech model based on Parler-TTS, offering high-quality voice synthesis with 2.33B parameters. Specializes in female voices with natural intonation.
Brief Details: ExLlama v2 quantized version of Mistral-Large-Instruct supporting 10 languages with multiple quantization options (2.2-6.5 bits/weight) for research use
Brief-details: Japanese text-to-speech model based on Parler-TTS, optimized for Japanese language with high-quality voice synthesis capability at 878M parameters. Beta version available for research and commercial use.
Brief Details: An 8B parameter LLaMA-3 based model fine-tuned on Orca agent instructions, optimized for conversational AI and text generation tasks using BF16 precision.
BRIEF DETAILS: Large 70B parameter Llama-based model missing LM head, released for research. Part of Tulu-3 family focused on instruction-following and diverse tasks.
Brief Details: An 8B parameter LLaMA-based model optimized for role-playing and storytelling, combining NIHAPPY's narrative capabilities with Mythorica's emotional depth and V-Blackroot's character consistency.
Brief Details: A specialized LoRA model trained on sketch card images, built on FLUX.1-dev base model. Optimized for 768x1024 resolution with constant LR scheduling and AdamW optimizer.
Brief-details: A powerful 72B parameter chat model quantized to 4.65 bits per weight, rivaling GPT-4 performance with support for 32K context and optimized for systems with 48GB VRAM
Brief Details: A 7B parameter GUI action model built on Qwen2-VL, specializing in generating thoughtful reasoning and executable actions for GUI tasks.
Brief-details: CraftsMan is a high-fidelity 3D mesh generation model that combines native 3D generation with interactive geometry refinement, supporting both English and Chinese inputs.
BRIEF-DETAILS: CamemBERTav2 is a powerful French language model with 111M parameters, trained on 275B tokens using DebertaV2 architecture, optimized for various NLP tasks.
Brief-details: Windows-compatible wheel distribution of flash-attention library, offering CUDA support and build tools for efficient attention mechanisms
Brief-details: A Vietnamese-English bilingual vision-language model with 938M parameters, optimized for document understanding and QA tasks using Colpali pipeline architecture.
BRIEF DETAILS: Stable Diffusion 3.5 Large GGUF - Advanced text-to-image Multimodal Diffusion Transformer with 13.9B params. Features improved image quality, typography & complex prompt understanding.
Brief Details: A 32B parameter AWQ-quantized code generation model with 128K context length, optimized for programming tasks and code reasoning. Based on Qwen2.5 architecture.
Brief-details: A 2-bit quantized version of Qwen2.5-72B-Instruct using AutoRound and GPTQ format, optimized for efficient deployment while maintaining performance.
