News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
imgPaliGemma

PaliGemma

A 3B open vision-language model built on SigLIP and Gemma-2B

PaliGemma is an open vision-language model that pairs a SigLIP-So400m image encoder with the Gemma-2B language model. It was presented in July 2024 as a broadly knowledgeable base model meant to be transferred to many tasks, and evaluated on almost 40 of them.

Share on WhatsApp

  • Open vision-language model combining SigLIP-So400m with Gemma-2B
  • Around 3 billion parameters, built as a base model for transfer
  • Evaluated on almost 40 tasks, including remote-sensing and segmentation
  • Paper first posted 10 July 2024, revised 10 October 2024

What it is

PaliGemma is an open vision-language model, meaning one system that reads both pictures and words and answers in text. It combines two parts: SigLIP-So400m, which turns images into representations the model can use, and Gemma-2B, the language model that produces the output. The team describes it as a versatile base model with broad knowledge, built to be adapted to new tasks rather than shipped as a finished assistant. It is around 3 billion parameters in size. A paper first posted in July 2024 reports tests on almost 40 different tasks, from standard vision-language benchmarks to narrower jobs such as reading remote-sensing imagery and segmentation.

How it came about

The model was introduced in a paper posted on 10 July 2024, with Lucas Beyer and Xiaohua Zhai among the 35 listed authors. A revised version followed on 10 October 2024, adding two appendices and a few citations but no change to the model itself. The work sits in the computer vision, artificial intelligence, computation and language, and machine learning research areas.

Inside the model

PaliGemma joins a vision encoder to a language model. The authors say it was trained as a versatile, broadly knowledgeable base model intended to transfer well, meaning it should adapt cleanly when fine-tuned on a specific job. They evaluate it across almost 40 tasks, mixing standard vision-language benchmarks with specialised ones such as remote-sensing and segmentation. The published summary does not give layer counts, training data volumes, tokenisers or compute figures.

What it can do

The team reports strong results across a wide range of open-world tasks. Testing covered almost 40 tasks, including standard vision-language benchmarks as well as more specialised work such as remote-sensing imagery and segmentation. The abstract presents these as transfer results from a base model rather than as a single headline score.

How to use it

PaliGemma is described as an open vision-language model. The paper is publicly available as arXiv:2407.07726.

Questions people ask about PaliGemma

What is PaliGemma?

It is an open vision-language model that takes in both images and text and produces text. It pairs the SigLIP-So400m vision encoder with the Gemma-2B language model, and is sized at roughly 3 billion parameters.

Who made it and when?

A team of 35 authors presented it in a paper first posted on 10 July 2024, with Lucas Beyer and Xiaohua Zhai among them. A revised version of the paper appeared on 10 October 2024.

What is it designed for?

It is meant as a versatile, broadly knowledgeable base model that transfers well to new tasks. The authors tested it on almost 40 tasks, from standard vision-language benchmarks to specialised work such as remote-sensing and segmentation.

How do I get it?

The model is described as open, and the accompanying paper is public as arXiv:2407.07726.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on PaliGemma yet, so the facts on this page come from the sources listed above: arxiv.org/abs/2407.07726. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.