Leading AI companies are using food-inspired codenames for their next-generation artificial intelligence models, revealing a quirky Silicon Valley tradition that extends into the competitive AI race. Meta has codenamed its future AI frontier model “Avocado,” according to a CNBC report, continuing the tech industry’s penchant for culinary nomenclature. This naming choice comes as CEO Mark Zuckerberg doubles down on AI investments, with Meta planning to spend more than $70 billion this year on AI infrastructure alone. This massive expenditure follows a $14 billion investment in Scale AI and the recruitment of its founder, Alexandr Wang.
OpenAI, the creator of ChatGPT, has dubbed its new large language model “Garlic,” The Information reports. The company appears to be “spicing things up” as it faces increased competition from Google’s recent advances and navigates through a series of internal challenges. Interestingly, Garlic is separate from another LLM OpenAI is developing, codenamed “Shallotpeat,” suggesting the company is working on multiple AI models simultaneously.
Google has taken a different approach by making its codename public. The tech giant’s AI image generator in Gemini is officially named Nano Banana Pro, released on November 20. While Google had internally called the model “nano-banana,” they initially kept this playful choice under wraps before deciding to embrace it publicly.
The tradition of food-related codenames has deep roots in AI development. OpenAI’s “Strawberry” codename for its o1 model remains one of the most memorable, likely referencing the viral challenge where AI models struggled to correctly count the number of R’s in the word “strawberry.” Before Strawberry, OpenAI worked on the mysterious “Q” project*. Earlier this year, Elon Musk’s xAI indulged its sweet tooth by codenamed an early testing version of Grok-3 “Chocolate.”
Not all companies follow the food theme. Mistral AI, the France-based startup, chose “Jaguar” for a testing model, while Anthropic opted for artistic inspiration with its model family names: Opus, Sonnet, and Haiku—three different types of literary compositions. This naming convention mirrors Google’s earlier Android operating system, which famously used dessert names for years before discontinuing the practice.
Key Quotes
Meta plans to spend more than $70 billion this year on AI infrastructure
This staggering investment figure demonstrates the massive financial commitment Meta is making to compete in the AI race, representing one of the largest single-year technology infrastructure investments in history and underscoring CEO Mark Zuckerberg’s all-in strategy on artificial intelligence.
OpenAI has hit a rough patch, feeling the heat from Google’s advances and stumbling with a series of missteps
This assessment reveals the competitive pressure facing OpenAI, the company that launched the AI boom with ChatGPT, showing that early leadership in AI doesn’t guarantee continued dominance as tech giants like Google rapidly advance their own capabilities.
Our Take
The food-themed codenames are more than Silicon Valley whimsy—they’re a window into the unprecedented AI arms race consuming tech’s biggest players. Meta’s $70 billion annual AI spend is particularly striking, representing roughly 10% of the entire company’s market value being deployed in a single year. This level of investment suggests these companies view AI not as an incremental improvement but as an existential imperative. The fact that OpenAI is developing multiple models simultaneously (Garlic and Shallotpeat) while facing competitive pressure indicates the breakneck pace required to stay relevant. What’s most telling is the shift from secretive development to public branding—companies are moving from research mode to market positioning, suggesting we’re entering a new phase where AI models become consumer-facing products with brand identities. The diversity of approaches, from Meta’s frontier models to Google’s image generators, shows the AI landscape fragmenting into specialized applications rather than a single general-purpose solution.
Why This Matters
These codenames reveal more than just Silicon Valley’s playful culture—they offer insights into the intense competition and massive investments driving the AI industry forward. Meta’s $70 billion infrastructure spend and $14 billion Scale AI investment demonstrate the astronomical costs of remaining competitive in the AI race. The fact that multiple companies are simultaneously developing numerous models (OpenAI’s Garlic and Shallotpeat, for example) shows the rapid pace of AI innovation and the pressure to continuously advance capabilities.
The shift from secretive codenames to public branding (like Google’s Nano Banana Pro) suggests companies are becoming more confident in their AI offerings and seeking to build brand recognition in an increasingly crowded market. OpenAI’s struggles mentioned in the article, coupled with Google’s advances, highlight how quickly leadership positions can shift in this fast-moving industry. For businesses and investors, these developments signal that AI infrastructure and model development require unprecedented capital commitments, potentially consolidating power among well-funded tech giants. The variety of approaches—from Meta’s frontier models to Google’s image generators—also shows how AI is diversifying across different applications and use cases, impacting everything from content creation to enterprise solutions.
Related Stories
- Big Tech’s 2025 AI Plans: Meta, Apple, Tesla, Google Unveil Roadmap
- OpenAI’s Competition Sparks Investor Anxiety Over Talent Retention at Microsoft, Meta, and Google
- Meta and Nvidia Billionaires’ Wealth Soars $152B in AI Boom
- OpenAI Lost Nearly Half of Its AI Safety Team, Ex-Researcher Says
- Google’s Sundar Pichai Responds to Microsoft CEO Satya Nadella’s ‘Dance’ Comment on AI in 2024
Source: https://www.businessinsider.com/ai-models-codenames-meta-avocado-openai-garlic-strawberry-2025-12