Open Vision-Language Models (VLMs)

Gemma is Google DeepMind’s family of open-weight AI models for developers building applications that can run on their own hardware. Its range covers compact models for phones and IoT devices alongside larger Gemma 4 models for reasoning on personal computers and servers. Some applications can work offline, keeping model inference on the device. Google AI Studio and Google Cloud are also available for hosted use.

Llama Guard is Meta's collection of downloadable AI content moderation models for developers building LLM applications. It checks user inputs and model responses for content that violates safety policies, including text and images. Developers can use the models in their own deployments or access moderation through Meta's hosted Llama API.

Qwen2.5-VL is a vision-language model you can run on your own hardware to answer questions about images and video. It's aimed at developers building document processing tools, visual assistants and agents that interact with computer or phone screens. The instruction-tuned 7B model has Apache 2.0 licensing and works with Hugging Face Transformers, with weights available in Safetensors format.

Pixtral is a family of Mistral models for developers who want to run multimodal AI on their own hardware. The associated mistral-inference project is archived and no longer maintained. That status applies to the inference library.

SmolVLM is a compact vision language model from Hugging Face for developers building local AI applications that work with images and text. It can describe pictures, answer questions about diagrams, and read information from documents such as invoices. Its small memory footprint makes on-device use practical on laptops and smaller local setups.

Nemotron is NVIDIA's family of AI models for developers building agents that reason, write code and call tools. You can run models locally for private, offline work or deploy them on your own servers. NVIDIA publishes model weights, training data and recipes so teams can inspect and adapt the models for their applications.

Mistral Small and Large are downloadable language models for developers building chat, reasoning and tool-using applications on their own infrastructure. Capabilities and hardware requirements depend on the release. Mistral Small 3.1 adds image understanding to text generation, while Mistral Large 2 is a larger text model.

MiniCPM-V is a family of local vision-language models for developers building apps that interpret images and video on their own hardware. It supports iOS, Android and HarmonyOS, as well as Mac deployment and server inference. The current repository states that MiniCPM-o/V code and model weights use Apache 2.0.

PaliGemma is a family of downloadable vision-language models for developers and researchers building applications that work with images and text. It combines SigLIP's image processing with Gemma's language capabilities to answer questions about visual content. Its main appeal is task-specific fine-tuning: you can adapt a base model to your own image data and intended use.

LLaVA is a family of vision-language models for researchers and developers who want to ask questions about images on their own hardware. It pairs a CLIP vision encoder with a language model to support image descriptions, visual reasoning and reading text in pictures. Its Python code is open source under Apache 2.0; the project places research-use restrictions on its data and checkpoints, with additional terms from the underlying models.

Granite is IBM's family of open-source AI models for developers and businesses that want to run and customize AI on their own hardware or servers. The language-model repository listed here is archived and no longer maintained. The broader family includes models for language, speech, document understanding and forecasting, released under Apache 2.0 for research and commercial use.

CLIP is an open-source image and text model that lets developers and researchers classify images using labels written in natural language, without collecting training examples for each task. It runs locally through PyTorch on a CPU or CUDA GPU. The code and model weights use the MIT license.

JoyCaption is an open-weight image captioning model for people preparing datasets to train or fine-tune diffusion models. It runs on your own GPU and covers both SFW and NSFW images, including photography, anime, digital art and furry artwork. Automated captions reduce the need to write descriptions by hand or find images that already have usable text.

Grounding DINO finds objects in images using category names or descriptive phrases you supply. It's a local AI model for developers and computer vision researchers who need detection beyond a fixed set of labels, including people building dataset annotation tools.

OpenCLIP is a Python and PyTorch library for developers and researchers who want to match images with text on their own hardware. It implements OpenAI's CLIP approach: images and descriptions become numerical representations that the model can compare. This supports image search and zero-shot classification, where text labels define the categories without a separate classifier trained for each task.

Janus-Pro is a multimodal AI model from DeepSeek that answers questions about images and creates pictures from text prompts. It runs on your own hardware and suits developers and researchers who want both capabilities in one model. A local Gradio demo provides a browser interface, while Hugging Face hosts a separate online demo.

GOT-OCR2.0 is an OCR model for developers and researchers who want to extract text from images on their own hardware. It handles both plain text and formatted output through a single model, with recognition modes for selected regions and documents spanning multiple pages. The Python codebase builds on Vary.

DeepSeek-OCR is an open-source OCR model for developers building document processing tools and researchers studying how AI reads text through images. It runs on your own hardware with NVIDIA CUDA GPUs. Its distinctive focus is visual text compression: representing document images with compact sets of vision tokens for a language model to read.

Molmo is Ai2's family of vision-language models, with code for running and training models on your own hardware. It's for developers and researchers who need to work with images and text, adapt a model, or evaluate it against visual tasks. The Python codebase is open source under Apache 2.0 and builds on OLMo, adding image encoding and generative evaluation.

Florence-2 is Microsoft's open-source vision model for developers who want to process images on their own hardware. It handles several image tasks through text prompts, so one model can generate descriptions, read text and locate objects. It runs locally with PyTorch and Hugging Face Transformers on a CPU or CUDA GPU, and uses the MIT license.

Moondream is a vision model for developers building software that needs to understand images. It can answer questions about a picture, write captions, locate objects, identify points and segment regions. The open-weight models can run on your own hardware, including in an air-gapped environment. The repository code is licensed under Apache 2.0; check each model checkpoint’s own terms for use.

Llama is Meta's family of large language models for developers, researchers and businesses that want to run models on their own hardware or servers. Its downloadable weights let you build generative AI applications with local inference. Access requires license acceptance and approval, and the weights use custom licensing for research and commercial use.

InternVL is a family of downloadable vision-language models for developers and researchers building AI that can interpret images and discuss them in text. It combines visual recognition with language models, supporting both multimodal chat and tasks such as image classification and image-text retrieval.

dots.ocr is a self-hosted document parser that combines multilingual text recognition and page layout analysis in one vision-language model. It's for developers and teams converting PDFs or document images into structured text while running inference on their own hardware. The Python project is open source under the MIT license.

Surya is a local OCR toolkit for developers extracting text and structure from PDFs and document images. It combines text recognition, layout analysis and table recognition in one vision-language model, so results retain page structure and reading order rather than just the words.

PaddleOCR is an open source OCR and document parsing toolkit for developers building document search, RAG systems and AI agents. It runs on your own hardware or a self-hosted server and turns PDFs and images into structured Markdown or JSON. The Python toolkit uses PaddlePaddle and carries the Apache 2.0 license.