Meta’s New AI Model Can Translate 200 Languages

July 7, 2022

Meta has open-sourced a new AI model called No Language Left Behind, which is capable of translating 200 languages, including 55 African languages.

While Meta is keen to use the new tool to improve its own products, open-sourcing will allow the model to be used by technologists to build other tools such as AI assistant.

Meta expects the model to support more than 25 billion translations across its own products every day. In addition to translating content and delivering better ads, the model will be used to detect harmful content and misinformation.

NLLB nearly doubles the number of languages covered by a single state-of-the-art AI model, and improves the quality of translations by an average of 44% compared to previous AI research.

Meta researchers built FLORES-200, a dataset that helps them evaluate NLLB performance in 40,000 different language directions. Meta also performed other automated metric evaluations, as well as human evaluations, to determine the quality of translations.

In support of the United Nations Sustainable Development Goals, Meta will award up to $200,000 in grants for effective use of NLLB-200 to researchers and non-profit organizations focused on sustainability, food security, and others.

The sources for this piece include an article in ZDNet.

Top Stories

Related Articles

December 31, 2025 Meta is buying Manus, a fast-growing agentic AI startup that already generates subscription revenue, in a deal more...

December 31, 2025 AST SpaceMobile has launched the largest satellite ever deployed in low-Earth orbit, escalating competition with SpaceX’s Starlink more...

December 31, 2025 Microsoft engineer Galen Hunt briefly set off alarm bells across the developer community after declaring an ambition more...

December 31, 2025 Global PC shipments could fall by as much as 9 per cent in 2026 as worsening memory more...

Jim Love

Jim is an author and podcast host with over 40 years in technology.

Share:
Facebook
Twitter
LinkedIn