BRIEF DETAILS: GGUF quantized version of Rin-9B model featuring multiple compression variants from 3.9GB to 18.6GB, optimized for efficiency and performance.
BRIEF-DETAILS: 14B parameter language model optimized for roleplay & Russian language support. Merged from Qwen 2.5 finetunes. Strong creative capabilities with stable performance.
BRIEF-DETAILS: Qwen2.5-3B merged model combining multiple Qwen variants using model stock method, built on Qwen2.5-3B-Instruct base for enhanced capabilities
Brief-details: FinBERT model fine-tuned for financial sentiment analysis with 88% accuracy. Classifies text as positive, negative, or neutral. Built on BERT architecture.
Brief Details: Specialized 300M parameter model fine-tuned for burn scar segmentation in satellite imagery, achieving 87.52% IoU for burned area detection using HLS data
Brief-details: A sophisticated 14B parameter merged LLM combining multiple high-quality models using SCE merge method, built on Qwen2.5 architecture with optimized performance and capabilities
Brief Details: A quantized T5-small model fine-tuned for grammar correction, achieving 0.88 BLEU score with FP16 optimization for efficient inference
Brief-details: A 7B parameter language model created through SLERP merging of pre-cursa-o1-v1.2 and post-cursa-o1 models, featuring optimized attention and MLP layer weights
Brief-details: An 8B parameter merged LLM combining Dolermed and Smarteaz variants of Llama 3.1, built on Dobby-Mini-Unhinged base using model stock merge method.
Brief-details: 8B parameter Llama 3.1-based merged model combining Smarteaz and Hermedive variants with Dobby-Mini-Unhinged base, using model_stock merge method
BRIEF-DETAILS: 8B parameter Llama 3.1-based merged model combining medical knowledge (MedIT-SUN) with DeepHermes capabilities, built using model_stock method
BRIEF-DETAILS: Ling-plus: A 290B parameter MoE LLM with 28.8B activated parameters, 64K context window, and open-source architecture optimized for scalability and adaptability.
Brief Details: Ling-plus-base is a 290B parameter MoE LLM with 28.8B activated parameters, featuring 64K context length and MIT license. Developed by InclusionAI.
Brief-details: Hindi sentence similarity model using SBERT architecture. Maps Hindi text to 768-dimensional vectors for semantic comparison and search.
Brief Details: A classifier model from ABSA trained for version 0.2, available on HuggingFace, designed for classification tasks with REST API compatibility.
Brief-details: Advanced PDF layout segmentation model built on LayoutLMv3, specialized for document structure analysis and block-level segmentation.
BRIEF-DETAILS: GLM-Edge-V-2B is a 2B parameter vision-language model from THUDM, capable of image understanding and text generation with bfloat16 support
Brief-details: Tiny-mixtral is a minimal test version of Mixtral, designed for CI/CD testing purposes. Not trained for production use. Created by TitanML.
Brief-details: Meta's Llama 3.3 70B instruction-tuned model optimized for 4-bit quantization, offering multilingual capabilities across 8 languages with 128k context window.
BRIEF DETAILS: Microsoft's WavLM-Base: Pre-trained speech model for full-stack audio processing, trained on 960h Librispeech data, optimized for 16kHz audio input.
BRIEF-DETAILS: Japanese-optimized DeepSeek-V3 variant with selective MoE layer experts, focused on Japanese language processing. GGUF format for efficient deployment.
