From Garlic to Avocado: The goofy AI model codenames you should know

In the ever-evolving world of artificial intelligence, leading tech companies have taken a playful approach to naming their projects, opting for food-inspired codenames that reflect both creativity and strategy. Recent reports reveal that prominent players like Meta, OpenAI, and Google have embraced this trend, creating a unique lexicon that adds a lighthearted twist to their serious endeavors. Meta has chosen to name its upcoming AI frontier model "avocado," as highlighted in a CNBC report. This decision comes as the company prepares to invest a staggering $70 billion in AI infrastructure this year, following a previous $14 billion investment in Scale AI. With CEO Mark Zuckerberg steering this ambitious pivot, it's clear that the costs associated with this venture are substantial. OpenAI, meanwhile, is navigating its own challenges, facing increased competition from Google. In response, the company has named its new large language model "garlic," according to The Information. This model is distinct from another one in development, which has the codename "Shallotpeat." These whimsical names reflect OpenAI's desire to inject some fun into its innovation process amidst a backdrop of intense competition. On the other hand, Google seems to have enjoyed its food-themed codenames so much that they've made one of them public. Their AI image generator, Gemini, is officially known as "Nano Banana Pro," a title that was initially kept under wraps. This playful moniker was revealed on November 20, showcasing Google's willingness to embrace a lighter side. The trend continues with OpenAI's earlier project named "strawberry," which cleverly referenced the challenges AI models face in accurately identifying the number of Rs in the fruit's name. Before "strawberry," OpenAI had a secretive project known as Q*. Adding to the mix, Elon Musk's xAI took a sweet approach with the codename "chocolate" for an early version of Grok-3, while French startup Mistral AI chose a more fierce name, "Jaguar," for its testing model. Additionally, Anthropic has introduced a family of models with poetic names like Opus, Sonnet, and Haiku, further emphasizing the creative naming conventions in the AI landscape.