Brief Details: Gemma 2B variant with 10M context length using recurrent local attention, optimized for <32GB memory usage. MIT-licensed with F32 tensor support.
Brief-details: AuraSR-v2 is a GAN-based super-resolution model for 4x upscaling of generated images, featuring 618M parameters and Apache 2.0 license
Brief Details: StarChat-Alpha: 15.5B parameter coding assistant model fine-tuned from StarCoder, optimized for programming tasks with FP16 precision
Brief-details: CodeGeeX4-ALL-9B is a powerful multilingual code generation model with 9.4B parameters, built on GLM-4-9B, offering superior code completion and generation capabilities.
Brief Details: Multilingual instruction-following LLM based on LLaMA 7B, supporting EN/ZH/JA/DE with enhanced role-playing and context handling capabilities.
Brief Details: OPT-13B-Erebus is a specialized text generation model based on OPT-13B architecture, trained on adult-themed content with 6 distinct datasets.
Brief Details: DeciLM-6b is a 5.7B parameter LLM optimized for efficiency, featuring variable GQA attention and 4096 token context window. Up to 15x faster than Llama 2 7B.
BRIEF-DETAILS: WizardLM-70B-V1.0 is a powerful large language model built on Llama 2, achieving impressive scores on MT-Bench (7.78) and AlpacaEval (92.91%), specialized in following complex instructions.
BRIEF DETAILS: ChatGLM2-6B-INT4 is a quantized bilingual LLM offering 42% faster inference than its predecessor, with 8K context length support and improved performance across multiple benchmarks.
Brief-details: Allegro is an advanced open-source text-to-video generation model with 2.8B parameters, capable of creating 6-second HD videos at 15 FPS from text prompts.
Brief Details: BitNet b1.58 3B - An efficient 3.32B parameter model trained on RedPajama dataset, achieving comparable performance to FP16 models with binary weights.
Brief-details: ChatGLM3-6B-32K: Advanced 6B parameter LLM optimized for long contexts up to 32K tokens, with enhanced position encoding and specialized long-text training capabilities.
Brief Details: A specialized text-to-image model for generating D&D character art, supporting 29 species and 15 classes with detailed fantasy styling.
Brief Details: Multimodal 8.46B parameter model combining vision and language capabilities, specialized in OCR and document understanding with Apache 2.0 license.
BRIEF-DETAILS: SillyTavern-Presets offers customizable roleplay configurations for LLM interactions, featuring optimized sampling parameters and character card templates.
Brief-details: Zero-shot text-to-speech model supporting 6 languages (en, zh, ko, ja, fr, de) with non-autoregressive architecture and masked generative codec transformer technology
Brief-details: JetMoE-8B is a cost-efficient 8.52B parameter MoE model achieving LLaMA2-7B performance, with only 2.2B active parameters and trained for $0.1M.
Brief-details: Optimized 7B parameter Mistral model fine-tuned on OpenOrca dataset, offering strong performance with various quantization options and ChatML format support
Brief Details: 8B parameter LLaMA-based model optimized for creative writing and assistant tasks, featuring improved narrative capabilities and instruction following
Brief-details: High-quality anime-style text-to-image model optimized for 768x768px generation, featuring enhanced photorealistic capabilities and specialized anime aesthetics.
Brief-details: DeciCoder-1b is a 1.1B parameter code generation model optimized for Python, Java, and JavaScript, featuring Grouped Query Attention and 2048-token context window.
