Sakana Fugu Ultra 1 1 Review Masterclass
A review of Sakana Fugu Ultra 1.1, an AI model that uses a council of multiple expert models working together to generate higher-quality outputs than single models like Claude Fable 5. The model excels at complex tasks like game development but takes approximately 25 minutes per generation and has regional availability restrictions.
Summary
The speaker presents a detailed masterclass on Sakana Fugu Ultra 1.1, a new Japanese AI model that uses a multi-expert orchestration system rather than relying on a single model. The core innovation is routing prompts to a council of up to three specialized AI experts that work independently and then merge their outputs into a single final answer.
The speaker demonstrates the model's capabilities through multiple game development test cases, comparing outputs against Claude Fable 5. Examples include open-world games, 3D games, flight simulators, and parachute games. Notably, all these outputs were generated on the first attempt without iterative refinement, though each generation takes approximately 25 minutes.
The speaker identifies a fundamental limitation of single-model AI systems called the "one brain ceiling problem," where a single model handles all aspects of a complex task (physics, graphics, controls, rendering) and can drop context or fail on certain components when the task becomes too large. Using multiple specialized models addresses this by distributing different aspects of the problem across different experts.
Benchmark comparisons show Sakana Fugu Ultra 1.1 outperforming Fable 5 in smoothness, graphical quality, control responsiveness, and overall functionality. The speaker emphasizes that the model is integrated into the Agent Operating System (AOS), allowing users to save outputs and organize them in a workspace without needing external APIs or additional orchestration layers.
Regional limitations are noted: the model appears blocked in the UK and EU, but works in Asia and the US. The speaker suggests Open Router as an alternative for accessing the model in restricted regions. Other similar multi-expert approaches are mentioned, including Fusion and Hermes Mixture of Experts.
The speaker concludes by promoting the AI Profit Boardroom community, which provides the complete Agent OS setup with daily tutorials, 205+ pages of testimonials, and weekly coaching calls. The system is framed as accessible even for non-technical users.
Key Insights
- Sakana Fugu Ultra 1.1 beats Claude Fable 5 on benchmarks by using a council of models that work together, with an orchestrator routing tasks to one to three expert models that merge their outputs rather than relying on a single model
- The model operates under a one-shot generation paradigm where each prompt takes approximately 15-25 minutes to generate, meaning all outputs are first-attempt generations without iterative back-and-forth refinement
- Single AI models face a fundamental 'one brain ceiling problem' where one brain handles all aspects of a complex task (physics, graphics, controls, rendering) and begins dropping context or failing components when the job becomes too large
- The speaker argues that every AI model wins somewhere and loses somewhere on benchmarks, so winning users run a stack of models and route jobs to whichever brain or team of brains best fits the specific task rather than picking one favorite model
- Sakana Fugu Ultra 1.1 is blocked or unavailable in the EU and UK, but the model can be accessed through Open Router's API for users in restricted regions
Topics
Transcript
[0:00] Today we're going to be looking at Sakana Fugu Ultra, which is a new model that's just dropped. This is from Japan. And if you look at the benchmarks here, Sakana Fugu Ultra 1.1 is actually beating Fable 5 on the benchmarks. Now, I've run several tests on it. I'll show you what they look like in a second. Um, but yeah, basically Sakana Fugu 1.1 just dropped and it's unlike any other AI model you've used because you can give [0:30] it one prompt and it sends a whole team of AI experts to build it for you. I actually put that team through the hardest game builds on my benchmark and it actually matched the best model…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Julian Goldie SEO
Google AI Studio Just Got A Huge Security Upgrade
Google is reportedly developing a new "Security Check" mode for AI Studio that would apply secure coding practices to AI-generated applications. The feature aims to automate security verification during app development, addressing common vulnerabilities like exposed keys, weak authentication, and data access issues before deployment.
New Gemini Skills Update Is WILD!
Google has launched Gemini Skills, a new feature allowing users to save custom instructions once and reuse them repeatedly by typing a slash command, replacing the discontinued Gems feature. Skills can include reference files, be combined together, and are built on an open file format (skill.md) that allows portability across different AI tools.
NEW Codex CLI Update Is ABSURD!
OpenAI's new Codex CLI update transforms terminal-based programming assistance by enabling multiple AI agents to work simultaneously on different tasks, visible in a full-screen control center interface. The update includes features like Agent View, work trees, voice control, and parallel task execution, fundamentally changing how AI assistance operates from sequential Q&A to autonomous team-based work management.
NEW Hermes Cloud + Agent OS is Crazy! 🤯
Hermes Cloud combined with Agent OS enables users to deploy AI agents in the cloud 24/7 using free models without API keys or local setup complexity. The system provides persistent memory, sandbox isolation, and integration with custom workflows like content research, competitor analysis, and news monitoring.
Manus 2.0 Just Changed AI Agents Forever
Manas 2.0 represents a fundamental architectural overhaul that enables users to create games, websites, videos, and automations through simple text requests, with the key innovation being editable components rather than fixed outputs. The platform uses a new Cascade engine that reduces token usage by 23% and improves task speed by 28%, while maintaining all project elements in one connected workspace.