Brief Details: Pix2struct-base is a 282M parameter image-to-text model for visual language understanding, supporting 5 languages with state-of-the-art performance.
Brief-details: A specialized text-to-image Stable Diffusion model fine-tuned on biriyani food images, offering precise generation of Indian cuisine visuals.
Brief Details: Mistral-NeMo-Minitron-8B-Instruct is an 8.41B parameter instruction-tuned LLM by NVIDIA, offering 8K context, optimized for conversation and RAG tasks.
BRIEF-DETAILS: Efficient non-autoregressive text-to-image model producing high-res images, trained on H100 GPUs. Apache 2.0 licensed with 5.3K+ downloads.
Brief Details: A specialized Stable Diffusion variant fine-tuned for image-to-image generation using CLIP image embeddings instead of text, enabling DALLE-2-like variations.
BRIEF DETAILS: GPT5o-reflexion: 6.74B param LLaMA-based model with perfect benchmark scores. Features reflection capabilities and specialized system prompt for complex reasoning.
Brief Details: A 7B parameter Mamba-architecture language model trained by TII, offering competitive performance without traditional attention mechanisms.
Brief Details: Large-scale image tagging model with 315M params, specializing in Danbooru content classification. Supports ratings, characters, and general tags using EVA02 architecture.
BRIEF DETAILS: Llama-3 based 8B parameter model optimized for roleplay, featuring Orthogonal Activation Steering and GGUF quantization. Balanced RP/non-RP training with improved response compliance.
BRIEF DETAILS: A 15.5B parameter code-generation model fine-tuned on CommitPackFT & OASST, supporting 80+ programming languages with strong multi-language performance.
BRIEF-DETAILS: 14B parameter multilingual LLM excelling in English, Chinese, Japanese and Korean. Strong performance in chat, long-context handling (up to 320k tokens), and RAG tasks.
Brief-details: GGUF-compatible Grok-1 quantized model with various compression options (Q2_K to Q6_K), optimized for llama.cpp with 316B parameters
Brief-details: 16ch-VAE is an open-source VAE reproduction for SD3, offering high-quality image encoding with 31.5151 PSNR and improved performance over SD1.5/SDXL VAEs.
Brief-details: An uncensored 13B parameter LLaMA-based model trained on Orca-style data, achieving 53.56% average on key benchmarks with strong performance in reasoning tasks.
BRIEF-DETAILS: A German-focused 7B parameter LLM based on Mistral, optimized for German language tasks while maintaining English capabilities. Features ChatML format and DPO training.
Brief Details: CogAgent-18B is a powerful visual language model with 18.3B parameters, specializing in GUI operations, visual dialogue, and high-resolution image processing up to 1120x1120.
Brief Details: OpenBuddy LLaMA2 13B is a multilingual chatbot supporting 8 languages, built on Meta's LLaMA architecture with 13B parameters for text generation tasks.
BRIEF-DETAILS: BiLLa-7B-SFT is a bilingual LLaMA variant optimized for Chinese-English reasoning, featuring enhanced language modeling and ChatGPT-generated analysis integration.
Brief-details: 4-bit quantized version of Falcon-7B-Instruct optimized for inference, featuring 1.54B parameters and GPTQ compression for efficient deployment.
Brief-details: Anime-style text-to-image model trained on 550k images, featuring high cosine similarity with ACertainty (85%). Built for high-quality animation generation.
Brief-details: A specialized text-to-image embedding model trained for SD 2.0, focusing on Egyptian sci-fi aesthetics with distinct orange/blue color schemes and stylized ancient Egyptian elements.
