# Multimodal processing

Gemma 3n-E2B-it is a lightweight open-source multimodal model launched by Google, built on the same technology as Gemini and optimized for low-resource devices.

Gemma 3n is a lightweight and state - of - the - art open - source model family launched by Google, supporting multimodal input and output.

Gemma 3n is a lightweight and state-of-the-art open-source multimodal model family launched by Google. It is built on the same research and technology as the Gemini model and supports text, audio, and visual inputs.

NuExtract 2.0 is a series of multimodal models specifically trained for structured information extraction tasks. It supports text and image inputs and has multilingual processing capabilities.

A quantized version of the image-to-text model based on Gemma 3 4B, aiming to make knowledge accessible to the public

Gemma 3 is a lightweight open-source multimodal model launched by Google, built on Gemini technology, supporting text and image input and generating text output.

This model is converted from ds4sd/SmolDocling-256M-preview to the MLX format, supporting image-text-to-text tasks.

Gemma 3 is a lightweight open model series launched by Google, built on the same research and technology as the Gemini model, supporting multimodal input and text output.

Gemma 3 is a series of lightweight open models launched by Google, supporting multimodal input (text and images), with a 128K large context window, suitable for various tasks such as question answering and summarization.

A document Q&A specialized model fine-tuned based on sberbank-ai/ruBert-large, supporting Russian and English document Q&A tasks.

A document QA model fine-tuned from sberbank-ai/ruBert-large, excelling at extracting answers from documents, supporting Russian and English.

A document question-answering model fine-tuned based on sberbank-ai/ruBert-base, excelling at extracting answers from document contexts, supporting Russian and English.

Ola-7B is a multimodal language model jointly developed by Tencent, Tsinghua University, and Nanyang Technological University, based on the Qwen2.5 architecture. It supports processing image, video, audio, and text inputs and outputs text.

This model converts PDF documents into Markdown format while preserving the original document layout structure and accurately recognizing mathematical formulas and tables.

A 4-bit quantized version based on the Mistral community's Pixtral-12B, focusing on image text-to-text tasks and supporting Chinese description generation.

This is a version of Microsoft's Florence-2 model fine-tuned for 1 day using the Docmatix dataset (5% of the data) with a learning rate of 1e-6

This is a multimodal model fine-tuned based on Microsoft's Kosmos-2 model, specifically designed for recognizing Pokemon names on Pokemon cards.

Cell segmentation model developed by Sribd-med team, suitable for cell instance segmentation tasks in multimodal images

A model based on the Donut architecture, specifically fine-tuned for Latvian receipt data

A speech-to-text transformer model trained for end-to-end English-to-German speech translation

A Transformer-based end-to-end speech translation model designed for English to Romanian speech translation

End-to-end English-to-French speech translation model based on S2T architecture, trained on the MuST-C dataset