Want to add the Hermes Mixture of Agents skill to your agent? Here’s how it upgrades your output — and what it technically is.
MoA gives your agent a panel of models instead of one, fusing their answers into a stronger result. It’s one of the simplest ways to level up an agent’s output.
Key takeaways
- MoA upgrades your agent by fusing several models into one stronger answer.
- Strictly, it’s a model provider, not a Hermes “skill” in the Skills-System sense.
- It beats any single model on hard tasks — and switches on with one command.
Is MoA A Hermes “Skill”? The Honest Answer
Hermes has a real Skills System, so it’s fair to ask if Mixture of Agents is one of those skills. The honest answer: not technically.
MoA is a virtual model provider — you select a preset as your model, and the panel runs inside the normal agent loop. So while it absolutely is a skill in the everyday sense of “a capability that makes your agent better”, it’s not a Skills-System skill you install. Knowing that saves you hunting in the wrong place.
How MoA Levels Up Your Agent
The upgrade is simple but powerful. Instead of your agent thinking with one model, several models each give their take privately, and an aggregator combines the best into the final answer.
It’s the difference between asking one expert and convening a panel. On hard tasks, the panel wins — your agent gets noticeably sharper answers without you changing anything else about your workflow.
Does It Actually Work?
On Hermes Bench, a two-model panel (Opus 4.8 aggregating over a GPT-5.5 reference) beats either model alone:
- Opus + GPT-5.5 panel: 0.8202
- Opus 4.8 alone: 0.7607
- GPT-5.5 alone: 0.7412
A panel of experts beats one genius — roughly 8% above Opus and 11% above GPT on hard tasks.
When This Capability Is Worth It
Like any upgrade, use it where it counts. MoA shines on complex reasoning and important work; for quick tasks a single model is faster and cheaper.
It does use more tokens (several models run per turn), but you can pair cheaper models and still beat one expensive model alone — strong value for hard problems.
How To Turn It On
- Run
hermes update - Run
hermes modeland pick the Mixture of Agents provider - Choose a preset (or set your own in
config.yaml) - Switch with
/model default --provider moa, or use/moafor a one-off
Where I Run It
I run MoA inside my Agent OS, next to Fusion and Sakana Fugu — three takes on the same panel-of-models idea, one click apart. Hermes only just shipped MoA, but I’ve run this pattern for weeks via those two.
Want the whole stack done for you with live coaching? It’s in my AI Profit Boardroom (3,800+ operators). New to Hermes? Start free with my AI Money Lab. And for the full overview, see my guide to Hermes Mixture of Agents.
What Makes MoA Such A Powerful Upgrade
Most ways to improve an agent are incremental. MoA is different because it changes the quality of every hard answer your agent gives, not just one part of the workflow.
By combining perspectives, it catches what a single model misses and smooths over individual weaknesses. That’s why a panel reliably beats even its strongest member on difficult tasks.
MoA vs Just Picking A Better Model
You might ask: why not just use the best single model? Because the best models keep getting gated, and even when you have one, a panel of models usually beats it on hard problems.
So instead of chasing one elusive top model, you upgrade your agent with a panel you can build today. It’s a more reliable way to level up than waiting for the next release.
Stop Chasing The Model, Build The System
This is the lesson I keep coming back to. The model is a swappable part; the system around it is what you actually own and improve.
MoA is a perfect example — a panel of current models outperforming a single frontier model you can’t even access. Build the system, and you stop being at the mercy of release schedules.
The Bottom Line
As a capability, Mixture of Agents genuinely levels up your agent — even if it’s technically a model provider rather than a Skills-System skill.
Switch it on for hard tasks, keep a single model for the quick stuff, and let the panel do the heavy lifting where it counts.
Getting Started Today
You can add this capability to your agent in a couple of minutes. Update Hermes, open the model picker, choose the Mixture of Agents provider and a preset, and you’re running a panel.
Start by pointing it at one genuinely hard task you’d normally struggle with, and compare the answer to a single model. That side-by-side is usually all it takes to see why the panel wins.
The One Thing To Remember
If you take one thing away, make it this: a panel beats a genius. Several good models, combined well, outperform any single model on the problems that matter.
That’s the whole reason this capability exists, and it’s why I keep it switched on for my hardest work.
Who Should Turn It On
If you push your agent on genuinely hard problems — complex builds, deep reasoning, important decisions — this capability is for you. If you mostly do quick tasks, you may rarely need it.
Either way, it costs nothing to set up, so it’s worth having ready for the moments that matter.
FAQ
Is Mixture of Agents a Hermes skill?
Not in the Skills-System sense — it’s a virtual model provider. But it is a capability that levels up your agent.
What does it do for my agent?
It fuses several models into one stronger answer on each turn.
Does it really make answers better?
Yes — a panel scored 0.82 on Hermes Bench vs 0.76 for Opus alone.
How do I add it?
Run hermes update, then hermes model, and select the Mixture of Agents provider.
Does it cost more?
More tokens per turn, but cheaper models combined can beat one expensive model.