Hermes agent feels internet explorer when compared to mercury agent.

So… reasoning is disabled in mercurys dsv4 flash? There was no reasoning done at all on the left. Is this a paid ad?

Wait, I just installed and tested, didn’t disable anything. Lets assume if reasoning was disabled, that won’t matter if the results are still like this for basic prompts on same machine with same AI provider. I mean I was expecting hermes to be a bit slow but this is something next level.

this true @NousResearch ?

this is you right now

I just tried this on my Hermes and it did not take that long at all, is this some sort of rigged partnership? Reasoning enabled btw… smh.

Do better, This is blatant bullshit. Turn reasoning on in Mercury and repost real results.

That's a little misleading. My Hermes agent does not take that long to initialize.

Nice try diddy

Another one: Told mercury and hermes to fetch latest tweet reply from @Teknium and yet hermes failed on this general task.

Vish! Falta anos de desenvolvimento pro mercury chegar ao nível do Hermes. Não adianta ser rápido e inútil.

unfollowed - obviously a paid partnership

A better test would be to give them both the same model as engine, same prompt, then compare their outcome, not their alleged speed. Performance means superior outcomes, not faster slop production.

You must have set Hermes Agent to “xhigh” and Mercury to no thinking because of course if they were the same models at the same reasoning mode setting they would have taken almost similar thinking time, so it just didn’t make sense what you tried to prove there.

Give your mercury agent memory a durability boost and dispatch task agents with context-packs

This is off brand.

I can also show you a raw api-calling script that uses flash models with reasoning off. It will suck hard but it will sure print text damn fast. I’ll probably use Hermes to build it. Did you also build Mercury using Hermes?

There is a reason picking the right harness matters :

hermes giderek aptallaşıyor gibi

🤝

What kind of slop is this dude trying to throw around? I'm going to do a comparison of these two and post what I find out. I've been using Hermes for a while and it's going to take a lot to get me to change. Way more than this.

That's not true Hermes has a feature where you can do a blank install and add everything à la carte. it's very modularized for this reason so you can have small compact builds

Memory bandwidth and model parameter size affect the speed more than an agentic process…

This is a paid demo! Shame

Another one:

Guys, chill out. I don't think it's malicious, he's obviously still figuring it out. @HuggingModels - the reason hermes is slow in your example is due to reasoning being enabled. If you disabled reasoning on hermes, you will get quick responses, like you did with mercury 👇

Enjoy the loss of subscribers.

I don't even give a fuck whether this is fake or not. Might try mercury because hermes is getting into the phase where every update there's some random bullshit to deal with. Where with openclaw it was just bugs, with hermes it's questionable decisions.

Mercury & Orgo = Money printer

👀 Lets see what this guy can do...

Please share the final results.

unfortunatetely Fable 5 ran into a block... this is what we got so far:

❤️‍🩹

2 Indians commie talk to each other

Meh, another slop generating, coattail-following, click bait artist making up bullsh!t on X. If it weren't for the few legit people on here, I'd be gone already. At least I can mute you though 👋

Embarrassing to post this

Speed or Quality?

if you want the same leap in features while keeping the speed try this

Now give them a complex task and show us the results.

Dude, turn on Reasoning. My GPT 5.5 Direct API and Codex think longer in some cases than actual Hermes 😂

not true at all, buggy since installation, need to make code edits to make it work, opened PR, but i dont care anymore, im uninstalling this crap. No place to configure telegram in Settings page. Yes, it feels like internet explorer, but IE5, from 1999