Sakana Fugu Ultra promises to deliver “the very best frontier-level performance” by… using frontier AI models. Japanese AI startup Sakana has released benchmark results for its Fugu models, claiming they perform comparably to or better than leading AI models such as Fable 5, Gemini 3.1 Pro, and GPT-5.5. The Fugu models are designed to strategically utilize other advanced AI models like Claude and Gemini for specific tasks, without disclosing which model is used. Sakana plans to enhance Fugu over time by integrating newer and more efficient models, including their own.

• Sakana’s Fugu AI models show competitive or superior performance compared to Fable 5, Gemini 3.1 Pro, and GPT-5.5.
• Fugu models strategically employ other advanced AI models like Claude and Gemini for specialized tasks.
• The specific AI model used for a task is not disclosed to the user.
• Sakana aims to continuously improve Fugu by incorporating newer and more efficient AI models. Continue reading https://www.theverge.com/ai-artificial-intelligence/953904/sakana-fugu-ai