BRIEF DETAILS: A 24B parameter Mistral-based model optimized for reasoning, coding, and general tasks. Features low-temperature operation and ChatML format support, trained on 800k reasoning traces.
BRIEF DETAILS: Advanced text-to-speech model extending LLaMA architecture with 3B parameters, trained on 250K hours of Chinese-English data. Supports both direct text and speech-prompted synthesis.
Brief Details: DeepSeek-VL2-Tiny is a 1.0B parameter MoE vision-language model optimized for visual understanding, offering strong performance in visual QA and OCR tasks.
Brief-details: Powerful 18.4B parameter MOE model combining 8 Llama-3.2 3B models. Specialized for creative writing, fiction & roleplay. Uncensored output with vivid prose capabilities.
Brief Details: FLUX.1-Redux-dev is an AI model by black-forest-labs, designed for non-commercial use with specific licensing requirements governed by FluxDev agreement.
Brief-details: FLUX.1-Fill-dev is a specialized development model from black-forest-labs, focused on fill operations with strict non-commercial licensing requirements.
Brief Details: Aya-expanse-8b is an 8 billion parameter language model from CohereForAI, part of the Aya model family focusing on expansive language capabilities.
BRIEF-DETAILS: T5-v1.1 XXL encoder converted to GGUF format, optimized for embedding generation and image processing tasks, supports Q5_K_M quantization
Brief-details: Advanced text embedding model built on Mistral-7B, achieving SOTA retrieval performance (60.2 on MTEB). Specialized in text similarity and semantic search tasks.
BRIEF DETAILS: Mistral-7B-Instruct-v0.1 is a 7B parameter instruction-tuned language model from MistralAI, optimized for following natural language instructions and conversational tasks.
BRIEF-DETAILS: 14B parameter LLM based on Qwen 2.5 architecture, optimized for reasoning and multilingual support. Features 128K context window and strong performance in structured tasks.
BRIEF DETAILS: A 3B parameter Japanese-English language model with strong performance in Japanese QA, math, and coding tasks. Trained on 10T tokens with 3-phase training approach.
BRIEF DETAILS: 14B parameter multilingual LLM based on Qwen 2.5 architecture, optimized for reasoning and long-context processing (128K tokens), supporting 29+ languages.
Brief-details: A 14B parameter LLM based on Qwen 2.5 architecture, optimized for reasoning and multilingual support with 128K context window. Strong performance in structured responses and chain-of-thought reasoning.
Brief-details: A specialized LoRA adapter for HunyuanVideo T2V model enabling keyframe-based video generation with precise control over start/end frames, optimized for human subjects at specific resolutions.
BRIEF-DETAILS: LLaDA-8B-Base is a groundbreaking 8B-parameter diffusion model trained from scratch, achieving performance comparable to LLaMA3 8B
Brief-details: SkyReels-V1-Hunyuan-I2V is a state-of-the-art human-centric video foundation model capable of generating high-quality cinematic videos with advanced facial animations and Hollywood-level aesthetics.
Brief-details: Tifa-DeepsexV2-7b: Advanced 7B parameter model based on Qwen2.5, featuring MGRPO algorithm, 1M token context, enhanced role-play capabilities and CoT reasoning.
Brief-details: Zonos-v0.1-transformer is a state-of-the-art multilingual text-to-speech model trained on 200k+ hours of speech data, offering voice cloning and emotional control capabilities.
Brief-details: Powerful 14B parameter LLM with 1M token context length, optimized for long-context tasks while maintaining short-task performance. Features GQA attention and custom vLLM deployment.
Brief Details: Lumina-Image-2.0 is a 2B parameter flow-based diffusion transformer for text-to-image generation, featuring CPU offloading and customizable inference parameters
