Claude Fable 5 VS Hermes moa VS Fusion: Who Wins?
The video compares three AI code generation systems—Claude Fable 5, Hermes Mixture of Agents, and Open Router's Fusion—by testing them across 42 different builds including games, web apps, and interactive projects. Fable 5 and Hermes Mixture of Agents tie in overall performance with 3 wins each, while Fusion underperforms, demonstrating that newer frontier models don't always outperform strategically combined older models.
Summary
The creator tests three AI systems side-by-side to determine which produces the best outputs for various coding tasks. Hermes Mixture of Agents works by combining multiple models together with a judge that fuses answers, while Fusion uses multiple agents working as a panel with a judge merging their outputs. Claude Fable 5 is Anthropic's newer frontier model tested against these alternatives.
Across the 42 test builds, results are mixed. In the first three tests (a game, Nordic Crypt game, and orbit visualization), Fable 5 wins 2 out of 3. The orbit test is notably won by Hermes Mixture of Agents, showing it can outperform the newer model. For landing page creation, Hermes Mixture of Agents produces the best output with superior UI design and animations, while Fable 5 struggles with UI layout. The Dragon Realm test results in a three-way tie.
A critical test involves a web OS with multiple applications. The Fusion version fails to autosave notes and has unresponsive elements. Hermes Mixture of Agents has nicer UI but exhibits buggy behavior. Fable 5 wins decisively—all applications work smoothly without bugs. In 3D environment tests (Voxil and Frost Fell), results are closer, with Fusion winning Voxil and the Frost Fell test being a tie. A Doom-style game test results in a win for Hermes Mixture of Agents because Fable 5 cannot move forward properly.
The creator reveals that Hermes Mixture of Agents performs competitively despite using older models because each output receives multiple perspectives from different models, with a judge evaluating and combining the best elements based on evidence. The final tally shows Fable 5 with 3 wins, Hermes Mixture of Agents with 3 wins, and Fusion with 0 wins. The creator recommends using Fable 5 as the default due to ease of use with CLI and workspace integration, but suggests Hermes Mixture of Agents as a token-efficient backup for longer prompts, especially after Fable 5's subscription availability ends July 7th.
Key Insights
- Hermes Mixture of Agents performs competitively against Fable 5 despite using older models because every draft gets quality control from a panel of agents with a judge that fuses answers based on evidence, selecting the best ideas from each model's output
- Fable 5 struggles specifically with UI design for landing pages and websites, producing messy layouts, while Hermes Mixture of Agents excels at creating visually appealing landing pages with smooth animations and glowing effects
- Fable 5 demonstrates superior functionality for complex interactive applications like a web OS, with all applications working smoothly without bugs, while both Fusion and Hermes Mixture of Agents produce buggy, unresponsive versions
- Fusion claims to beat Fable 5 on benchmarks but does not demonstrate that superiority in practical testing, winning zero rounds across the 42 test builds while both Fable 5 and Hermes tie with 3 wins each
- Using Hermes Mixture of Agents as a backup to Fable 5 is strategically valuable because Fable 5 subscription users will run out of tokens quickly, and Hermes can handle longer 10-15 minute prompts more cost-efficiently
Topics
Transcript
[0:00] So, Fable 5 versus Hermes, Mixture of Agents versus Open Routters Fusion. Who wins? Today, we've tested them out. I'm going to show you the results in a second. This is something we actually built with Fable 5 a few minutes ago. So, let's get straight into this. We've compared them side by side. I'm going to show you exactly what the difference is. Now, if you're wondering, okay, what is Hermes mixture of agents? This is a way of combining multiple models together to get better outputs you'd normally get with, for example, something that's frontier. So, for example, if you wanted [0:30] to create outputs that are better than Opus 4.8, you could combine GPT 5.5 and…
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.