Sakana AI has announced the launch of Fugu Ultra, a frontier-level orchestration model designed to rival top-tier AI models such as Anthropic's Fable 5 and Mythos Preview. This new release is available to the general public worldwide through a single, OpenAI-compatible API, making it accessible for both individual and enterprise customers. Fugu Ultra is tailored for users with demanding, multi-step tasks across engineering, science, research, cybersecurity, and data analysis, addressing the need for resilient AI infrastructure that avoids single-vendor risks or export control restrictions.

Fugu Ultra distinguishes itself by utilizing a multi-agent orchestration system: it is a language model that can autonomously delegate subtasks to a pool of expert LLMs, including instances of itself, choosing and combining their outputs for the highest quality results. Benchmark comparisons show that Fugu Ultra matches or surpasses leading competitors on coding, reasoning, and scientific tasks. Early user feedback highlights its strengths in sustained persona consistency and thoroughness in long, complex workflows, such as code review and security assessment, where it surfaces more issues and maintains focus better than other models.

Sakana AI positions this release as a step toward AI sovereignty, offering ongoing improvements as new models are added to the agent pool. The company, established with a mission to build collaborative AI ecosystems rather than monolithic models, emphasizes Fugu Ultra’s role in providing organizations with greater operational and geopolitical security.

Source