Brief-details: GGUF-converted version of Stable Diffusion 3.5 Large for efficient text-to-image generation, optimized for 8.15B parameters
BRIEF-DETAILS: Hybrid SSM-transformer 1.2B parameter model combining Mamba and transformer architectures. Features low latency, small memory footprint, and competitive performance.
BRIEF-DETAILS: Specialized ControlNet model for Flux.1-dev that processes surface normal maps, enabling precise control over 3D structure in image-to-image generation.
Brief-details: A specialized LoRA model for generating cute cat images in "eric cat style", built on FLUX.1-dev base model. Features text-to-image capabilities with specific trigger words.
Brief-details: Phantasma Anime is a Stable Diffusion LoRA model specializing in upbeat anime-style illustrations with fantasy elements and FX details, built on SDXL base.
Brief Details: AuraFace-v1: Advanced face recognition model using Resnet100 architecture with ArcFace. Achieves 99.65% accuracy on LFW benchmark. Apache 2.0 licensed.
Brief-details: xLAM-7b-fc-r is a 6.91B parameter function-calling optimized LLM from Salesforce, designed for efficient tool use and API interactions with competitive accuracy of 88.24% on BFCL benchmark
Brief Details: First instruction-tuned Gemma-2-9b model optimized for Chinese & English users. 9.24B parameters, using ORPO fine-tuning and flash-attn-2.
Brief Details: OpenELM-270M is a 272M parameter efficient language model by Apple, featuring layer-wise scaling and comprehensive pre-training on 1.8T tokens.
Brief-details: PuLID is a NeurIPS 2024 accepted model for pure and lightning ID customization using contrastive alignment, with FLUX and SDXL variants.
Brief-details: 8B-parameter Mamba-2-Hybrid language model combining Mamba-2, attention, and MLP layers, trained on 3.5T tokens with 4K sequence length context.
Brief-details: A fine-tuned 8B parameter LLaMA-3 model optimized for function calling and JSON mode outputs, featuring BF16 precision and conversational capabilities.
BRIEF DETAILS: A 9.39B parameter MoE-adapted Mistral model with 16 experts, trained on OpenHermes-2.5. Optimized for efficient instruction following and conversational tasks.
Brief-details: Dolphin 2.6 Phi-2 GGUF - A 2.78B parameter conversational AI model based on Microsoft's Phi-2, optimized for chat with multiple quantization options
Brief Details: A powerful 12B parameter multimodal LLM combining EVA02-5B and Zephyr-7B-β, excelling in visual-language tasks with RLHF alignment and real-time interaction capabilities.
Brief Details: A fine-tuned 7B parameter LLM based on Mistral, optimized for scientific reasoning with 32k context window and trained on 1B+ tokens.
Brief Details: T2I-Adapter for SDXL that enables lineart-guided image generation, featuring 77M parameters and Apache 2.0 license. Built by TencentARC.
Brief Details: A 3.6B parameter Japanese GPT-NeoX model fine-tuned with RLHF using PPO, optimized for instruction-following and conversation. Shows 47% improvement over SFT variant.
Brief Details: A specialized 349M parameter vision-encoder-decoder model fine-tuned from nougat-base, optimized for converting mathematical equation images to LaTeX code with high accuracy.
Brief Details: Meta's LLaMA 7B GGML - Efficient CPU+GPU inference model with multiple quantization options (2-8 bit). Optimized for performance and memory usage.
Brief Details: ChatGLM-6B-INT8 is a quantized bilingual LLM with 6B parameters, optimized for Chinese-English dialogue, running on 8GB RAM with INT8 precision
