AI tools for Multimodal

Inkling 2 Small is Thinking Machines Lab's multimodal language model for general assistance, coding and tool-using applications. It accepts text and images and generates text, w

NitroFusion2 is a text-to-image checkpoint designed to render a result in one or two diffusion steps. It builds on Z-Image-Turbo and targets sharp visual detail, including text

Ling-3.2-Flash is Ant Group's hosted multimodal model for reasoning over text, images and video. Its model overview describes applications in visual understanding, spatial inter

SenseNova 6.8 Flash is SenseTime's multimodal agent model for office work and content creation. It supports tasks that span several steps or sessions and can deliver outputs suc

Gemini Omni 1.2 Flash generates and edits short audiovisual clips from text and visual references. The model supports up to 15 seconds of generated video with synchronized nativ

LongCat-2.5 is Meituan's hosted multimodal model for extended agentic work. It combines image understanding with code generation for tasks such as working across complex project

Tencent Hy4 is the final hosted release of Tencent's flagship model for long-running agent tasks. It adds visual understanding to the earlier preview's text capabilities and sup

SenseNova-Loop-9B-Preview is an open multimodal language model for long-context and agentic workflows. It accepts text and images and generates text, supporting document interpr

LFM3 is Liquid AI's family of compact models for specialized and on-device intelligence. Its first release includes text, vision, audio, omni, and decision variants built around

Gemini 4 Carbon is Google DeepMind's frontier model for sustained engineering, reasoning, and multimodal work. The initial rollout focuses on selected cyber defenders and truste

Groot is an open-weight model family for reasoning and tool-using workloads. The release offers several size and precision variants, including the compact Groot-19B-A3B model, s

Nemotron 3.5 Super VL is NVIDIA's customizable multimodal model for specialized enterprise agents. Its 120B-parameter mixture-of-experts architecture activates 12B parameters pe

Muse Spark Open is Meta's open-weight multimodal model for long-horizon coding and agentic reasoning. Its mixture-of-experts architecture has 816B total and 42B active parameter

Unpaired Rosetta is a cross-modal embedding-alignment implementation that does not require paired examples. Wasserstein Procrustes alignment estimates an orthogonal map between

Mistral Large 4 is Mistral's trillion-parameter multimodal model for engineering, enterprise work, and complex agents. Its mixture-of-experts architecture activates 52B paramete

Nano Banana 2.1 is Google's Gemini-based model for image generation, editing, and visually grounded text responses. Based on Gemini 3.6 Flash, it accepts images and text in a co

EmbeddingGemma 2 is Google's lightweight model for representing multiple content modalities in one embedding space. It extends text embeddings to combinations of images, audio,

PixelUMM is a unified model for understanding and generating images and video directly in pixel space. A single decoder-only transformer processes image patches and video patche

Gemini 4 Argon is Google's frontier model for complex software engineering and professional knowledge work. Long-context reasoning supports multi-step tasks across large evidenc

MiMo-V2.6 is Xiaomi's natively omnimodal model family for coding, agent workflows, and visual problem solving. Pro and Flash variants are trained with scaled reinforcement learn