Meta readies Mango and Avocado AI models to challenge rivals
TL;DR
Meta is reportedly developing an image- and video-focused AI model codenamed Mango, a coding-oriented large language model called Avocado, and early-stage 'world models.' The plans were discussed internally by Meta’s chief AI officer Alexandr Wang in a Q&A with product chief Chris Cox, with both Mango and Avocado targeting a first-half 2026 release.
About this summary
This article aggregates reporting from 4 news sources. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
Race to AGI Analysis
Mango and Avocado are Meta’s clearest signal yet that it intends to be in the top tier of foundation model labs, not just a fast follower. Mango targets image and video generation, a space where OpenAI’s Sora and Google’s Nano Banana and Veo have raised the bar, while Avocado aims squarely at high-end code generation. The fact that both are being developed inside a centralized Meta Superintelligence Labs unit shows the company is willing to reorganize deeply around frontier AI, poaching talent from OpenAI and other labs to do it.
The more interesting piece for the AGI race is Meta’s renewed focus on “world models” – systems that build internal simulations of the environment to reason about actions over time. That’s exactly the direction many researchers think you need for generally capable agents and robotics. If Meta can marry powerful visual models (Mango), strong coding ability (Avocado), and accurate world models, it starts to look less like a social-media company and more like a full-spectrum AGI contender. For OpenAI and Google, this raises the stakes: the arms race is no longer just about benchmarks, but about who can operationalize rich multimodal understanding fastest.
