Meta Delays Launch of Avocado AI Model Following Performance Shortcomings Against Google Gemini and Other Rivals
Meta has postponed the rollout of its next-generation AI model codenamed 'Avocado' to May 2026 after internal tests revealed it underperforms compared to leading AI models from Google, OpenAI, and Anthropic.
Key takeaways · 5
- 01
Meta's Avocado AI model launch is delayed to at least May 2026 due to underperformance in internal evaluations against Google Gemini 3.0, OpenAI, and Anthropic models.
- 02
Avocado outperforms Meta's previous Llama 4 and Google's Gemini 2.5 but lags behind the latest state-of-the-art systems in reasoning, coding, and writing tasks.
- 03
Meta is considering licensing Google's Gemini technology temporarily, marking a rare admission that its AI capabilities are not yet competitive enough.
- 04
Despite unprecedented AI infrastructure investments estimated between $115-135 billion in 2026 and the formation of a 100-person elite 'TBD Lab,' Meta is facing execution challenges.
- 05
The company is moving away from its earlier open-source approach with models like Llama, leaning toward keeping new AI model code private.
Background: Meta's Avocado AI Project
Meta has invested heavily into its AI ambitions, assembling the elite TBD Lab led by Chief AI Officer Alexandr Wang. Avocado forms the flagship of a new fruit-themed AI model suite aimed at surpassing competitors.
Performance Evaluation and Delay
Internal testing showed Avocado outperforming older models but failing against current-leading models such as Gemini 3.0 in key areas including logical reasoning and coding. This shortfall delayed the launch from March to at least May 2026.
Strategic Considerations and Industry Impact
The potential licensing of Google's Gemini engine for Meta's consumer AI products is a strategic pivot that could shift competitive dynamics in Silicon Valley. Meta's move toward proprietary code also signals changes in AI development openness.
Financial and Organizational Implications
Meta’s enormous AI infrastructure spending in 2026 – nearly doubling 2025’s outlay – coincides with significant layoffs (20% workforce reduction) to balance costs amid product shortcomings.
Meta’s struggle to deliver a competitive AI model despite record investments reveals that leadership in AI requires more than capital—it demands innovation, execution excellence, and strategic flexibility. The potential of licensing a competitor's AI model marks a significant industry inflection point that may reshape competitive dynamics in AI-powered consumer technology, impacting billions of users and influencing the future direction of AI development and deployment.
Why it matters
Put this to work — one session a day, built for your industry.
Create a free account for a daily session — eight questions and one real-work challenge, on the news that affects your role.
Start freeSources
- Meta Delays Avocado AI Model to May After Internal Testing Reveals Performance Gaps with Google, OpenAI and Anthropic — The Machine Heraldmachineherald.io
- Meta Delays 'Avocado' AI Model After Google Gemini Crushes Internal Testsaidevdayindia.org
- Meta May Delay ‘Avocado’ As Tech Underwhelms In Tests Against Google Gemini And Anthropic AI Models: Report – The Business Investorthebusinessinvestor.co.uk
- Meta Avocado AI Delayed: $135B Bet Trails Gemini 3.0 | byteiotabyteiota.com
- Nvidia Becomes a Major Model Maker With Nemotron 3 - WIREDwired.com
- Meta delays rollout of new AI model after performance concernsbusinesstimes.com.sg
- Meta Delays Rollout of New AI Model After Performance Concernstech.slashdot.org
- Meta delays new AI model rollout after performance concerns - MSNmsn.com