Archive — US — techcrunch
Meta is developing a new image and video model for a 2026 release, report says
Meta is developing new AI models called 'Mango' for images/videos and 'Avocado' for text, aimed for a 2026 release. Led by Alexandr Wang, the models target advancements in coding and visual understanding. Amid restructuring and competition with AI leaders, Meta needs to deliver impactful products amid leadership changes and researcher turnover.
Key points
- Meta is working on two AI models: 'Mango' for images/videos and 'Avocado' for text.
- Both models are planned for release in the first half of 2026.
- The projects are part of Meta's Superintelligence Lab led by Alexandr Wang.
- The text model 'Avocado' aims to enhance coding capabilities and reasoning with visual information.
- Meta is currently lagging behind competitors like OpenAI, Anthropic, and Google in the AI sector.
- The AI division has undergone significant restructuring, including leadership changes.
- Some newly hired researchers have already left Meta, indicating challenges in talent retention.
Why it matters
- The shift in Meta's AI strategy reflects the broader trend in the tech industry toward advanced AI capabilities, especially in competition for market leadership.
- 2025 IT Trends emphasize the importance of strong AI models and infrastructure to enable businesses to adapt to changing consumer demands and regulatory environments.
- Meta's restructuring is a response to ongoing competition and could lead to innovative breakthroughs or further setbacks in AI development.
Meta's development of 'Mango' and 'Avocado' highlights its bid to regain competitive ground in AI, but success hinges on overcoming internal challenges and delivering standout products by 2026.