Brief Details: NVIDIA's 70B parameter LLaMA 3.3 model optimized with FP4 quantization, offering 3.3x memory reduction while maintaining 92-95% accuracy on benchmarks
Brief-details: MiniMax-Text-01 is a 456B parameter LLM featuring hybrid attention mechanisms, 1M token training context, and state-of-the-art performance across various benchmarks.
BRIEF-DETAILS: 405B parameter LLaMA-based model optimized for instruction following, math, coding & reasoning tasks. Strong safety measures & comprehensive benchmarks.
BRIEF DETAILS: 4-bit quantized version of Microsoft's Phi-4 (14B) model optimized by Unsloth, offering 70% reduced memory usage and 2x faster inference
Brief-details: AIN is a breakthrough Arabic-English bilingual multimodal model trained on 3.6M samples, excelling in OCR, medical imaging, and cultural understanding tasks with SOTA performance
Brief-details: A comprehensive collection of roleplay-focused presets (Methception, LLamaception, Qwenception) optimized for different AI models, featuring enhanced continuity and immersive storytelling capabilities.
BRIEF-DETAILS: Adventure-focused 12B parameter LLM optimized for challenging gameplay, trained on Nemo base model with pessimistic sentiment and high-stakes narratives
Brief-details: Stability AI's 3D-aware model for point cloud processing and generation, offering advanced capabilities in 3D scene understanding and manipulation.
Brief-details: Medical LLM built on Qwen2.5-7B, specialized in complex medical reasoning with bilingual capabilities (EN/CN) and unique thinking-before-answering approach
Brief-details: Advanced 8B-parameter multimodal LLM with Mixed Preference Optimization, combining vision and language capabilities using ViT-MLP-LLM architecture for enhanced reasoning
Brief-details: Advanced 78B parameter multimodal LLM combining Qwen2.5-72B-Instruct and InternViT-6B vision model, optimized with Mixed Preference Optimization (MPO) for enhanced reasoning
Brief Details: A 70B parameter LLM fine-tuned for tool usage and multi-turn dialogue, achieving SOTA performance on Berkeley Function-Calling Leaderboard.
Brief-details: A versatile 12B parameter LLM built on Mistral-Nemo-Base, featuring comprehensive capabilities in writing, roleplay, analysis & tool use with 32K context.
Brief-details: A 2.21B parameter vision-language model optimized for OCR, image analysis, and math problem-solving with multilingual support and video understanding capabilities.
BRIEF-DETAILS: Clarity is WeMake's specialized orchestration AI that combines multimodal processing with organizational intelligence, supporting both German and English languages.
Brief-details: ModernBERT-large is a 395M parameter BERT-style model with 28 layers, trained on 2T tokens of text/code data, supporting 8,192 token context length.
Brief-details: MMAudio is an advanced AI model for video-to-audio synthesis, focusing on high-quality sound generation through multimodal joint training techniques.
BRIEF DETAILS: 30M parameter dense embedding model by IBM, producing 384-dim vectors. Optimized for English text similarity and retrieval with Apache 2.0 license. Fast performance with 49.1% MTEB score.
BRIEF-DETAILS: A 12B parameter merged LLM combining UnslopNemo and Mag-Mell models using SLERP method. Optimized for coherent outputs with ChatML template support and various GGUF quantizations available.
Brief Details: General OCR Theory model capable of processing various text formats including documents, tables, formulas & sheet music. Built on unified end-to-end architecture.
Brief-details: FLUX.1-Canny-dev is a specialized AI model from black-forest-labs focused on edge detection and image processing capabilities, requiring non-commercial licensing.
