I Tested the Fable 5 Killer (Hermes Agent)

Name: I Tested the Fable 5 Killer (Hermes Agent)
Uploaded: 2026-06-24T17:33:10.235Z
Duration: 19 min 6 s
Channel: Jack Roberts
Description: A practical comparison of two AI models (Sakana Fugu and GLM 5.2) against Claude Opus 4.8 within an agent framework, featuring tests on tool-calling, website-building, and coding.

Jack Roberts24 June 2026Watch on YouTube

Part of series

Ep. 6 · Fugu Ultra: Multi-Agent AI

Diepgaande verkenning van Sakana AI's Fugu Ultra als parallel multi-agent systeem dat meerdere modellen simultaan inzet.

View the series

Description

📈 ALL Systems: https://bit.ly/4kol0y5 🩵 Free Resources: https://bit.ly/3RNNDLa 🔥 Type with your voice: https://bit.ly/4eGoI3R Two models have been claiming to kill Claude Fable 5 in the last week, so I put both through real tests inside Hermes agent. Sakana Fugu is an intelligent router that secretly orchestrates a pool of frontier models behind one API, while GLM 5.2 is an open workhorse from China at a sixth of the cost. I run them against Opus 4.8 on tool calling, building a one page website, and improving the Hermes operating system. The results genuinely surprised me, especially how GLM 5.2 held up. I also walk through the exact model routing strategy I'm using going forward. *Core Software* 🤖 Hermes Agent 🐟 Sakana Fugu 🧠 GLM 5.2 ☁️ Claude: https://claude.ai/ 🐳 DeepSeek: https://www.deepseek.com/ 🔌 Zapier MCP 🔥 Glaido: https://bit.ly/4eGoI3R ⌚️ Stamps: 00:00 - Two Models Killed Fable? 00:36 - The Game Of Thrones Problem 01:48 - How Sakana Fugu Works 03:01 - GLM 5.2 The Open Workhorse 03:48 - Connecting Sakana To Hermes 04:46 - Testing Sakana Out 05:40 - Connecting GLM 5.2 06:30 - Checking GLM Works 07:15 - Test One Tool Calling 09:07 - Tool Calling Results 10:05 - Test Two One Page Website 11:01 - Comparing The Websites 13:53 - Test Three Hardest Yet 15:16 - The Build Results 16:09 - Fugu Is Not A Fable Killer 17:02 - The Routing Strategy I Use 18:33 - Why GLM 5.2 Impressed Me I tested Sakana Fugu and GLM 5.2 against Claude Opus 4.8 inside the Hermes agent to see which model is actually best for tool calling, building websites and coding, plus how to connect them with Open Router and Zapier MCP. #HermesAgent #GLM52 #SakanaFugu #ClaudeOpus #AIAgents #AIModels

What you'll learn

Sakana Fugu is an intelligent router orchestrating a pool of frontier models behind one API, rather than a standalone base model.
GLM 5.2 performs surprisingly well for its cost (one-sixth of Claude Opus) on practical tasks like tool-calling and coding.
In real agent tests against Claude Opus 4.8, Sakana Fugu did not prove itself a true 'Fable Killer' despite the hype.
The test tasks were concrete: tool-calling, building a one-page website, and improving the Hermes operating system.

Frequently asked questions

What is Sakana Fugu and how does it differ from GLM 5.2?

Sakana Fugu is an intelligent router that orchestrates a pool of frontier models behind one API. GLM 5.2 is an open-source workhorse model from China that functions as a standalone solution at a much lower cost.

Did Sakana Fugu beat Claude Opus 4.8 in the practical tests?

No, despite recent claims to kill Claude Fable 5, Sakana Fugu did not prove itself as a true killer against Claude Opus 4.8 in real agent tests.

How is GLM 5.2 connected to Hermes agent and what does it cost?

GLM 5.2 can be connected to Hermes agent via Open Router. The model costs roughly one-sixth of Claude Opus, significantly reducing your expenses.

What three tests did the creator run to compare the models?

The three tests were: tool-calling, building a one-page website, and improving the Hermes operating system.