Creating a New Fable? Testing Mixture of Agents in Hermes Agent

Nous Research's new Mixture of Agents combines several models into one — they claim it beats Opus 4.8 and GPT-5.5. So I tested how much it actually costs, how slow it is, and whether the quality is worth it. Sign up for my FREE weekly newsletter, where I spill my unfiltered thoughts on the latest AI news, cool research, and projects I'm building: https://www.onchainaigarage.com/ 🐦 Follow Tonbi on X for real-time AI x blockchain updates! https://x.com/tonbistudio Hermes Agent's new Mixture of Agents (MoA) lets you pair several "reference" models that quietly advise a single "aggregator" model that does the actual tool calls and answers — Nous Research claims an 8% edge over Opus 4.8 and 11% over GPT-5.5 on their upcoming Hermes Bench. In this video I explain how MoA works under the hood, set it up and build custom presets in the dashboard and desktop app, then put it to a non-scientific test on the questions everyone asked after my short demo: cost, speed, and quality. I run a browser-mini-game build as a single model versus an MoA of open-weights models (GLM 5.2, Kimi K2.6, MiniMax M3), then a second round pitting GPT-5.5 solo against GPT-5.5 advised by a "three-headed Grok." The whole point is to see whether combining models actually buys you better output — and what it costs you in time and money to get there. Resources: 🔗 Hermes Agent: https://github.com/NousResearch/herme... Timestamps: 0:00 - Mixture of Agents: Nous Research's new feature 2:01 - How MoA works: reference models + an aggregator 3:01 - Setting it up + building presets 5:04 - The test prompt: a browser mini-game 6:39 - Baseline: a single model (GLM 5.2) 9:30 - MoA round 1: three open-weights models 13:55 - Round 2: GPT-5.5 + a "three-headed Grok" 16:56 - Building the Grok-advised preset 18:07 - Final thoughts Have you tried Mixture of Agents yet, and what model combinations would you want to test? Let me know in the comments. If you want more hands-on agent feature breakdowns like this, please like, subscribe, and hit the bell! 🦐✨ #HermesAgent #MixtureOfAgents #AIAgents #NousResearch #LLM #GLM #Grok #AICoding #AITools