It may sound like a trip through the produce aisle, but leading AI companies have something much more important on their lists.
OpenAI, Meta, and Google have all relied on food-related names for their sometimes secretive plans for future AI models. OpenAI is even naming a highly anticipated advanced chip after a pepper.
Thinking with your stomach is nothing new for Silicon Valley, just look at the assortment of desserts Android assembled over the years before Google had its fill.
Here is a look at the mouthwatering and just plain goofy names AI and tech companies are using.
OpenAI: Jalapeño
When you're challenging Nvidia, the situation is bound to be spicy.
On June 24, OpenAI announced that the first advanced chip it developed with Broadcom is called "Jalapeño."
"Chips are foundational to the AI economy," OpenAI said in a statement. "Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI."
We’ve designed and built our first AI chip: Jalapeño.
— OpenAI (@OpenAI) Designed from the ground up by OpenAI and brought to production with[@Broadcom], Jalapeño is purpose-built for the LLM workloads powering ChatGPT, Codex, the API, and future agentic products.
Chips are foundational to the AI…[pic.twitter.com/mHU7DaMMTi][June 24, 2026] Garlic
The ChatGPT maker codenamed its new large language model "garlic," according to The Information. Garlic is separate from another LLM OpenAI is developing, codenamed "Shallotpeat."
Google: Nano Banana
Google appears to have loved a codename so much that it made it public. Google's AI image generator in Gemini is named Nano Banana Pro, which it released on November 20, 2025. Before then, Google had internally called the model nano-banana, though they had not publicly disclosed their zany choice.
Meta: Avocado
Meta codenamed its future AI frontier model "avocado", per a CNBC report. CEO Mark Zuckerberg didn't stick with the food theme.
Meta SuperIntelligence Labs dubbed the publicly released model "Muse Spark." Muse Spark was the first model released by Meta since the social networking giant spent billions overhauling its AI strategy, including a $14 billion investment in Scale AI to hire its cofounder, Alexandr Wang, to lead the division.
Past codenames branch out from fruit and veggies
The clearance section offers a wide selection of great names.
OpenAI might have one of the best all-time codenames with "strawberry," which it used to refer to its o1 model. The name was likely a play on the viral struggle of AI models to correctly identify the number of Rs in the fruit. Before Strawberry, OpenAI had a secretive project named Q*.
In 2025, Elon Musk's xAI had a sweet tooth when it codenamed an early testing version of Grok-3 "chocolate."
Mistral AI, the France-based startup, went in a completely opposite direction with "Jaguar," its codename for a testing model.
And Anthropic named its family of models Opus, Sonnett, and Hakiu, a trio of three different types of compositions. Recently, they've added Mythos and Fable, though the public release of those models has been complicated.