Brief Details: BioMistral-Clinical-7B is a 7.24B parameter clinical language model based on Mistral, optimized for medical and biological text generation with MIT license.
Brief Details: Moonshine-base: A 103M parameter ASR model for English speech recognition, optimized for resource-constrained platforms with MIT license.
Brief-details: SmolLM2-1.7B-Instruct-GGUF is a compact 1.7B parameter instruction-tuned language model available in multiple GGUF quantizations for efficient deployment.
Brief Details: SmolLM2-360M-GGUF is a compact 362M parameter language model optimized for efficient text generation, available in multiple GGUF quantization formats.
Brief Details: A 12.2B parameter Mistral-based language model fine-tuned on GU_instruct-Remastered-1.1 dataset, optimized for text completion tasks
Brief-details: SmolLM-1.7B-GGUF is a compact 1.7B parameter language model available in multiple quantization formats, optimized for efficient text generation using the GGUF format.
Brief Details: A powerful 91.9B parameter merged LLM based on Llama-3.1-Nemotron-70B, optimized using passthrough merging for enhanced instruction following capabilities.
Brief Details: 29B parameter creative writing model optimized for vivid storytelling with enhanced reasoning via Brainstorm 40x process. Features 128k context, uncensored output, and strong instruction following.
Brief Details: A specialized ASR model combining Whisper with NER capabilities - 1.54B params, supports speech transcription and entity recognition with optional masking.
Brief Details: A 14.8B parameter multilingual LLM optimized for English and German, featuring DPO fine-tuning and enhanced function calling capabilities
Brief Details: A large 92B parameter instruction-tuned language model merged from Llama-3.1-Nemotron-70B using mergekit's passthrough method, optimized for BF16 inference.
Brief-details: 8B parameter Japanese-enhanced LLM built on Llama 3.1, optimized for both Japanese and English tasks with strong instruction-following capabilities
Brief Details: MM-Embed is NVIDIA's state-of-the-art multimodal retrieval model that extends NV-Embed-v1, achieving 52.7 score on UniIR benchmark with text-image capabilities
BRIEF-DETAILS: A 3.4B parameter TIES-merged Qwen2.5 model combining MiniMix and RP-Mix variants, optimized for conversational AI using mergekit framework.
Brief-details: Model2Vec-based static embeddings model with 7.56M params, distilled from BGE base model. Fast compute, optimized for resource-limited scenarios.
Brief-details: A 22B parameter iMatrix-quantized GGUF model offering various quantization levels for efficient deployment, optimized for conversational tasks with enhanced performance.
Brief-details: A French language model trained on historical newspapers using ELECTRA architecture and TEAMS approach, optimized for Named Entity Recognition with 408GB training data.
BRIEF DETAILS: A 12.2B parameter merged model combining Starcannon v3 and NemoMix Unleashed, optimized for character roleplay with extended context handling and improved coherence.
Brief-details: A specialized LoRA model for generating high-quality mockup textures, trained on FLUX.1-dev base model with 22 hi-res images using AdamW optimizer and constant LR scheduler.
Brief Details: Arcee-VyLinh is a 3B parameter Vietnamese language model built on Qwen2.5, offering strong instruction-following capabilities with 32K context length.
Brief Details: A GGUF-converted version of Stability AI's SD 3.5 Medium model, optimized for image generation with 2.47B parameters. Requires ComfyUI-GGUF.
