# Multimodal Interaction

Moondream is a lightweight vision-language model designed for efficient cross-platform deployment. The 4-bit quantized version released on April 14, 2025 significantly reduces memory usage while maintaining high accuracy.

AgentCPM-GUI is an on-device graphical interface agent with RFT-enhanced reasoning capabilities, capable of operating Chinese and English applications, built upon the 8-billion-parameter MiniCPM-V.

UI-TARS-1.5-7B-4bit is a multimodal model focused on image-text-to-text conversion tasks, supporting the English language.

This is an MLX-format model converted from the Google Gemma 3-12B model, supporting image-text-to-text tasks.

VideoChat-R1-thinking_7B is a multimodal model based on Qwen2.5-VL-7B-Instruct, focusing on video-text-to-text tasks.

A vision-language-action model specifically designed for Minecraft, capable of executing thousands of in-game skills based on human language commands

UI-R1 is a vision-language model enhanced by reinforcement learning for GUI agent action prediction, built upon Qwen2.5-VL-3B-Instruct.

Vamba is a hybrid Mamba-Transformer architecture that achieves efficient long video understanding through cross-attention layers and Mamba-2 modules.

This is a video-text-to-text model based on the MLX format, developed by HuggingFaceTB, supporting English language processing.

Ultravox is a multimodal voice large language model built on Llama3.1-8B-Instruct and whisper-large-v3-turbo, capable of processing both voice and text inputs.

Fluxi AI is a multimodal intelligent assistant based on Qwen2-VL-7B-Instruct, capable of processing text, images, and videos, with special optimization for Portuguese language support.

UGround is a powerful GUI visual positioning model trained using a simple method, jointly developed by OSUNLP and Orby AI.

An intelligent vision-language model fine-tuned from HuggingFaceTB/SmolVLM-Instruct, optimized for training speed using Unsloth and TRL libraries

Dallah is an advanced multimodal large language model specifically designed for Arabic, with a focus on understanding and generating content across Arabic dialects.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting efficient segmentation through prompts.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting efficient segmentation through prompts.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting universal segmentation tasks through prompts.

The LLaVA-Video model is a 7B-parameter multimodal model based on the Qwen2 language model, specializing in video understanding tasks and supporting 64-frame video input.

xGen-MM is a series of the latest foundational large multimodal models (LMMs) developed by Salesforce AI Research, building upon the successful design of the BLIP series with foundational enhancements to ensure a more robust and superior model foundation.

A foundational model developed by FAIR for solving promptable visual segmentation tasks in images and videos

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting efficient segmentation through prompts.

A foundational model for promptable visual segmentation in images and videos developed by FAIR

UGround is a powerful GUI visual positioning model trained with a streamlined recipe, developed by the Ohio State University NLP Group in collaboration with Orby AI.

InternVideo2-Chat-8B is a video understanding model that combines a large language model (LLM) with video BLIP, built through a progressive learning scheme, capable of video semantic understanding and human-computer interaction.

LLaVA-MORE is an enhanced version based on the LLaVA architecture, integrating LLaMA 3.1 as the language model, focusing on image-to-text tasks.

An AI role-playing assistant based on the Llama 3 8B model, focused on creating immersive narrative experiences

An AI role-playing assistant based on the Llama 3 8B model, focused on creating interactive narrative experiences

InstructBLIP is a vision-instruction-tuned version based on BLIP-2, combining visual and language processing capabilities to generate responses based on images and textual instructions.

InstructBLIP is a vision instruction tuning model based on BLIP-2, using Flan-T5-xl as the language model, capable of generating descriptions based on images and text instructions.

InstructBLIP is the vision-instruction-tuned version of BLIP-2, combining vision and language models to generate descriptions or answer questions based on images and text instructions.

IDEFICS-9B is a 9-billion-parameter multimodal model capable of processing both image and text inputs to generate text outputs. It is an open-source replication of Deepmind's Flamingo model.