The strongest models are gated and access is granted only to a select few. Hermes Agent now exposes MoA presets as virtual models, giving you capabilities beyond the publicly available frontier: 8% higher than Opus 4.8 and 11% higher than GPT 5.5 on our upcoming benchmark.

Jun 26, 2026 · 8:49 PM UTC

325
593
6,211
1,833,538
HermesBench full leaderboard coming soon. Stay tuned!
21
20
465
50,734
Sort replies: Relevant Recent Liked
Replying to @NousResearch
Good stuff Hermes. Helping us normal folk to get more power <3
1
71
16,244
We aim to empower
2
5
130
15,091
Replying to @NousResearch
Now give me easy cloud access with a nice interface and take my goddman money!!
8
141
21,884
Very soon!
7
155
17,676
Replying to @NousResearch
is @OpenRouter behind this, right
2
12
19,582
Provider agnostic
3
2
102
18,765
Replying to @NousResearch
Gated access is temporary.
5
5
13,569
Can't do much with a "temporarily" gated model can you
3
106
12,721
Replying to @NousResearch
Very good! Will this benchmark be available for us to test our own moas?
1
3
2,957
Replying to @NousResearch
We are working on benching various combos of open source models to see if we can get Opus levels with much cheaper models as well, stay tuned!
99
44
1,337
134,879
Replying to @NousResearch
feel bad for people stuck using gpt-5.5 and opus-4.8 when we've got hermes over here
1
1
57
3,599
Replying to @NousResearch
same energy
2
34
2,644
Replying to @NousResearch
What about if models where never gated again. What about if you could give your Hermes a P2P inference network with access to any available model and confidential computation? Halo
3
2
21
2,277
Replying to @NousResearch
can I get early access?
2
15
1,922
Replying to @NousResearch
keep a watch on my token bills big bro
1
11
447
Replying to @NousResearch
anthropic is old news
We made ChatGPT, Grok, and Claude compete with our own model. They lost badly. The difference is our data: the largest set of expertly labeled human follicles in history. Models are a commodity now. Labeled biology is not. The task: trace all six layers of a hair follicle, pixel by pixel. ChatGPT and Grok never drew a map. Claude tried and scored 0.12. FOWLER scored 0.88. Same images. Seven times the accuracy. Same architectures as the big labs. Across 43 follicles it had never seen, FOWLER held at 0.73. Whoever owns the best data wins.
6
191
Replying to @NousResearch
are we still not allowed to auth with claude for opus 4.8? api key only?
7
392
Replying to @NousResearch
people clowned on fugu, but im interested
1
8
507
Replying to @NousResearch
Aura
6
117
Replying to @NousResearch
can I mux providers in MoA 👀
2
6
656
Replying to @NousResearch
YOU GUYS ARE F*CKING AMAZING
4
573
Replying to @NousResearch
我认为这是一项伟大的行动。
最强模型都被大公司锁着,只有少数人能用。 Hermes Agent 现在把 MoA 做成了虚拟模型,在他们 HermesBench 上比 Opus 4.8 强 8%、比 GPT 5.5 强 11%。 原理很简单:先让多个顶级参考模型并行思考(只给对话内容,不给工具),再由一个 aggregator(汇总模型)综合它们的结果去输出答案并调用工具。 不是自己从头训新大模型,而是把现有强模型聪明地组合起来用。 想自己体验?装个 Hermes Agent 后直接输入 /moa 就能切换(推荐先跑 hermes setup --portal 一键搞定)。 普通人也能用到这种 frontier 级能力,开源社区这波操作真的香! 感谢@NousResearch团队持续推进 🚀 #HermesAgent #MoA #开源AI
4
2,274
Replying to @NousResearch
Nous is one of those labs that do not waste their time on podcasts because they are busy shipping
1
4
638
Replying to @NousResearch
Definitely going to need to give this a try
3
457
Replying to @NousResearch
so hermes is basically the secret sauce huh
3
62
Replying to @NousResearch
"on our upcoming benchmark" ಠ_ಠ
2
320
Let’s go! This is the way forward!
2
66
Love this!
2
1,076
Replying to @NousResearch
nice, i also did fusion in openpaths too and integrated into hermes agent in a branch if yall are that way inclined :)
1
290