DeepMind tests "instant-ramen" Nano Banana model

Google DeepMind is accelerating its model optimization process by leveraging public model arenas to collect human-preference data on its upcoming lightweight models.

• –**Codenaming Strategy**: The name "instant-ramen" continues Google's practice of utilizing playful, food-themed codenames (such as "nano-banana") to benchmark unreleased models anonymously.
• –**Focus on Speed**: The "instant-ramen" moniker likely hints at exceptionally fast inference speeds or lower latency, matching the design goals of the Nano Banana line.
• –**Crowdsourced Benchmarking**: Testing in anonymous public arenas allows DeepMind to evaluate the model's performance against leading competitors before committing to a formal launch.