Definition
Google MUM (Multitask Unified Model) is an AI model presented by Google in May 2021. It is designed to answer complex queries requiring deep real-world understanding. MUM is 1000 times more powerful than BERT and stands out for its multimodal capability: it simultaneously understands text and images across 75 different languages. This enables it to synthesize information from multiple and multilingual sources to provide comprehensive answers to complex questions. MUM is gradually being integrated into various Google products, including Search, Lens, and rich results.
Key Points
- 1000 times more powerful than BERT according to Google
- Understands text and images across 75 languages simultaneously
- Synthesizes information from multilingual and multimodal sources
Practical Examples
Complex multilingual query
For the question 'I climbed Mont Blanc, how do I prepare for Kilimanjaro', MUM can combine sources in French, English, and Swahili to compare both ascents and provide a comprehensive answer.
Visual search
By submitting a photo of hiking boots to Google Lens, MUM can understand the image and suggest whether they are suitable for a specific trek.
Frequently Asked Questions
No, MUM does not directly replace BERT. Both coexist within Google's ecosystem. MUM is gradually deployed for specific use cases where its multimodal and multilingual power provides real added value.
Invest in in-depth, multi-format content (text, images, video), structure your data with Schema.org markup, and cover topics comprehensively so Google can extract rich information from your pages.
Go Further with LemmiLink
Discover how LemmiLink can help you put these SEO concepts into practice.
Last updated: 2026-02-07