Brief Details: CodeGen2.5-7B-multi is a powerful 7B parameter code generation model trained on StarCoderData, specializing in multi-language programming synthesis with infill capabilities.
Brief-details: A specialized Stable Diffusion model fine-tuned for China-Chic illustration style, combining traditional Chinese culture with modern aesthetics. Created by tilake with DreamBooth technology.
BRIEF DETAILS: ControlNet model for human pose-based image generation, trained on 200k pose-image pairs using SD-1.5. Enables precise control over human poses in generated images.
Brief Details: OPT-30B is Meta AI's open-source 30B parameter language model, designed for text generation and research accessibility, trained on 180B tokens.
Brief-details: EchoMimic is a state-of-the-art AI model for creating lifelike audio-driven portrait animations with editable landmark conditioning, supporting multiple languages and singing.
Brief-details: CogVideoX1.5-5B-SAT is an advanced open-source video generation model supporting 10-second videos with flexible resolution for both image-to-video and text-to-video tasks.
Brief-details: OpenELM-270M-Instruct is a 272M parameter instruction-tuned language model from Apple, offering efficient parameter allocation and strong zero-shot performance
Brief Details: MiniCPM-V is a 3.43B parameter bilingual visual-language model offering GPT-4V level performance, optimized for efficient deployment on various devices including mobile phones.
Brief Details: A 6.7B parameter code generation model built on DeepSeek-Coder, featuring execution feedback and refinement capabilities with strong benchmark performance.
Brief Details: DeciLM-6b-instruct is a 5.72B parameter instruction-tuned LLM optimized for speed, achieving 43.43% on ARC Challenge and featuring variable Grouped-Query Attention.
BRIEF-DETAILS: A powerful 72B parameter language model achieving state-of-the-art performance on multiple benchmarks, featuring innovative SGD dataset creation and advanced DPO learning techniques.
Brief-details: A powerful 46.7B parameter Mixtral-based model optimized for coding and conversational tasks, with GGUF format supporting various quantization levels for efficient deployment
Brief-details: AI router model that intelligently directs queries between GPT-3.5 and GPT-4, trained on extensive cross-domain benchmarks. Apache 2.0 licensed.
Brief-details: A 70.6B parameter LLaMA-3-based model optimized for function calling, achieving GPT-4 level performance (0.81 vs 0.80) with parallel function support and lower cost.
BRIEF-DETAILS: InternLM-chat-20b: A 20B parameter LLM trained on 2.3T tokens, featuring strong reasoning capabilities, 16k context length, and excellent performance across language, knowledge, and technical tasks.
Brief-details: InstantIR is a cutting-edge single-image restoration model that resurrects damaged images with extreme-quality details, supporting text-guided enhancement and customized editing.
BRIEF DETAILS: 104B parameter multilingual AI model with RAG & tool use capabilities. Offers various GGUF quantizations for efficient deployment & impressive perplexity scores.
🤖 Vicuna-13B v1.1: An open-source chatbot based on LLaMA, fine-tuned on 70K ShareGPT conversations. Features improved tokenization and loss computation.
Brief-details: Luna AI Llama2 Uncensored GGML is a CPU/GPU-optimized uncensored chatbot model with multiple quantization options (2-8 bit) offering flexible deployment and performance tradeoffs
Brief-details: Zephyr-7B is a fine-tuned 7B parameter LLM based on Mistral, optimized with DPO training on UltraChat and UltraFeedback datasets, offering strong chat capabilities.
Brief-details: Text-to-music AI model based on BART-base, fine-tuned on 282,870 text-music pairs to generate ABC notation sheet music from text descriptions.
