
Fugu Ultra v2 is Sakana AI's high-capability flagship model in the Fugu lineup, designed to achieve maximum performance on demanding and intricate assignments. It accepts text, images, and files to return written answers, making it practical for real-world jobs like reading an entire book-length document in a single pass, carrying out end-to-end software development, and conducting visual reasoning across charts and structured data. Rather than jumping straight to a final response, it can reason step by step, browse the web, connect to external tools, and organize its output into clean formats like JSON.
It is built for developers and organizations handling complex, multi-step tasks, autonomous research, and engineering workloads where output quality takes absolute priority. Unlike its sibling Fugu Max, which focuses on delivering the lowest possible cost, Fugu Ultra v2 pushes capability to the highest level by coordinating a flexible pool of open and specialized models rather than relying on individual third-party systems.