BRIEF-DETAILS: Mistral-7B-v0.1 is a powerful 7B parameter language model from MistralAI, featuring advanced architecture and strong performance across diverse NLP tasks.
Brief-details: Babel-9B-Chat is a multilingual LLM supporting 25 languages covering 90% of global speakers, with strong performance in reasoning, understanding, and translation tasks.
Brief Details: IBM's 2B parameter instruction-tuned LLM with enhanced reasoning capabilities, supporting 12 languages and specialized for controllable thinking tasks.
Brief-details: Zeta is a fine-tuned version of Qwen2.5-Coder-7B optimized for code edit prediction in Zed, featuring FP8 quantization and NGram speculative decoding
Brief-details: ReaderLM-v2: A 1.5B parameter LLM specialized in HTML-to-markdown/JSON conversion with 512K context window, supporting 29 languages and achieving 0.84 ROUGE-L score.
Brief-details: An 8B parameter multimodal LLM with GPT-4V-level capabilities for vision, speech and streaming. Supports real-time audio/video processing and voice cloning.
Brief Details: Flux text encoders - Specialized checkpoints for DualClipLoader in ComfyUI, enabling enhanced text encoding capabilities for AI image generation workflows.
Brief-details: A specialized 32B parameter LLM fine-tuned from Qwen 2.5 for enhanced deductive reasoning, particularly excelling at temporal reasoning tasks.
Brief Details: DeepScaleR-1.5B-Preview: A 1.5B parameter LLM fine-tuned via distributed RL, achieving 43.1% Pass@1 on AIME 2024, surpassing larger models through innovative scaling techniques.
Brief-details: Qwen2.5-VL-72B-Instruct is a powerful vision-language model with enhanced video understanding, agent capabilities, and visual localization features supporting 1-hour+ video analysis.
Brief-details: ONNX-optimized version of Phi-4 multimodal model, quantized to int4 precision for faster inference. Handles text, images, audio with 128K context window.
Brief-details: A specialized image generation model trained for creating cozy, TOK-style illustrations, featuring capabilities for scenes like characters in nature, animals, and fantasy elements. Requires "in the style of TOK" trigger phrase.
Brief-details: ALLaM-7B-Instruct is a bilingual Arabic-English LLM with 7B parameters, trained on 5.2T tokens and optimized for Arabic language tasks while maintaining English capabilities
Brief-details: Phi-4-mini-instruct-GGUF is a 3.8B parameter instruction-tuned model optimized for reasoning and math, supporting 128K context and 200K vocabulary across 23 languages.
Brief Details: State-of-the-art code embedding model (1.5B params) optimized for code retrieval tasks. Supports 9 programming languages with 1536-dim embeddings & 32k context.
Brief Details: A 2B parameter vision-language model optimized for document understanding, achieving SOTA performance on DocVQA (0.89) and ChartQA (0.87). Apache 2.0 licensed.
Brief-details: SmolVLM2-2.2B is a lightweight multimodal model for video/image analysis, requiring only 5.2GB GPU RAM. Excels at video understanding, image QA, and text transcription tasks.
Brief-details: Meta's 8B parameter LLM from the Llama 3 series - designed for efficient language understanding and generation with improved performance over previous versions
BRIEF-DETAILS: Advanced 12B parameter hybrid SSM-Transformer model with 256K context length, outperforming other open models in long-context tasks and RAG workflows.
Brief-details: A 7B parameter distilled model from DeepSeek-R1, based on Qwen2.5-Math-7B, optimized for mathematical reasoning and coding tasks.
Brief-details: GGUF-quantized video generation model optimized for ComfyUI, specialized in creating dynamic video content with custom prompts and workflows
