Hermes Agent For Marketing + Which AI Models Fit Landing Pages, Promo Videos And Weekly Briefs: Real GoldieBench Scores

Hermes Agent for marketing lets each job run on its own AI model, so the real question is which model to pin to the landing-page job, the promo job and the weekly brief.

I read the scores below on the live GoldieBench pages on 7 November 2026 and matched them to the marketing jobs Hermes can schedule.

The examples are a florist, a gym, a café and a small agency, and each pick says what that model does in the job.

Want my team to choose models for your own marketing jobs? Book a free SEO strategy session.

What GoldieBench Scores Tell A Marketer

The methodology page says every model gets the same fixed prompt, one shot, with no iteration and no best-of-several picking.

The output has to be one self-contained HTML file that runs in a browser, and Julian scores it from 0 to 10.

Two tasks map straight onto marketing work: Landing, a modern marketing landing page, and Aipbpromo, a motion-graphics video advert built as a web page.

Nothing on the bench scores captions, emails or research briefs, so for those jobs the overall average is only a rough guide.

Where Hermes Lets You Pin A Model

• The main model: set with hermes model, used by any job without its own pin.
• Each cron job: the cron docs allow a per-job pin with --model and --provider, or --pin to lock the current main model.
• The whole cron fleet: cron.model sets a default for every unpinned job, separate from your chat model.
• Each Bot: the Bot Mode docs give every Bot its own model, memory and skills.
• Thinking level: a job can pin its own reasoning effort with --reasoning-effort.

So a florist's landing-page job and her Monday brief never have to share one model.

The Shortlist With Live Averages

Averages, task counts and gold medals come from the all-models page.

Landing Pages: The Florist's Campaign Page

The Landing task page shows 26 models tested, 21 scored and a task average of 8.11.

A florist planning a seasonal page would give Hermes the offer, the delivery area and the brand skill, and ask for one HTML draft.

The job pins one of the 9.0 models, and the owner or her web person edits the result before it goes anywhere near the site.

Gemini 3.6 Flash at 8.6 is a fair pick when she wants several quick variations to compare.

Promo Videos: The Gym's Membership Push

The Aipbpromo task page asks for a cinematic advert with auto-playing scenes, animated stats and an end call to action, all in one HTML file.

It is harder: 26 models tested, 20 scored, and a task average of 7.48.

A gym wanting an animated web promo for a class launch would pin GPT-5.6 Sol or Opus 5 to that single job.

The Qwen 3.8 page says it was benched through Alibaba's Qoder platform, so check your Hermes provider actually offers it before you pin it.

Notice that Fusion and Grok drop from 9.0 on Landing to 7.5 and 7.0 here, so one task's winner isn't automatically the other's.

Weekly Briefs: The Café's High-Volume Lane

Most marketing jobs aren't one-shot builds; they're the same research-and-draft job, every week.

The Gemini 3.6 Flash page lists "high-volume agentic work where token cost dominates" under Best for, and suggests routing the everyday 90% to it.

A café's Monday topic brief and Wednesday caption drafts are exactly that kind of job.

Set it once as cron.model, and every unpinned job uses it while your chat model stays free for harder work.

Its 7.08 average is the lowest on this shortlist, which is fine for a list of topics and not for a page that has to sell.

Social Listening On X: Where Grok Fits

The Grok model page lists "workflows that need live X / Twitter context" first under Best for.

On the Hermes side, the providers page supports xAI with an API key or a SuperGrok or X Premium+ sign-in, and says the same sign-in is reused by X Search.

A small agency tracking what people say about a client's niche on X would pin Grok to that weekly listening job only.

Its 9.0 on Landing makes it a sensible second job too, if the agency wants one model for research and pages.

The Agency Room: One Model Per Bot

Hermes Bot Mode seats two to six Bots in a group chat, and each Bot keeps its own model.

For an agency's content room, this is the split I'd try, based on the scores above:

1. Researcher Bot: Gemini 3.6 Flash for the volume of weekly briefs.
2. Page-builder Bot: a Landing 9.0 model such as Fusion or Grok for campaign pages.
3. Editor Bot: Claude Opus 5, whose page lists the hardest agentic builds under Best for, at 8.27 overall.

The v0.23.0 release notes let a lead bot pick who answers, so the Editor can sit in that seat.

Kimi K3's page lists long-horizon agent runs under Best for, which suits a Bot that works through a long content calendar.

Thinking Levels: Light Briefs, Careful Pages

The model is one dial; the thinking level is the second.

The cron docs let a job pin its own reasoning effort, from none and minimal up to high and beyond, without touching your global default.

The same page suggests heavy scheduled analyses at high and cheap recurring jobs at minimal.

For marketing, that means the café's caption drafts can run light while the florist's campaign page and the agency's monthly report run with more thinking.

Gemini 3.6 Flash's page also lists fast prototype builds you iterate on, which fits a light, repeatable job better than a one-shot showpiece.

What The Scores Don't Change

No model choice adds an approval step for outgoing posts.

The Hermes security docs cover risky commands, so every job that reaches the public still needs "draft only" in its wording, whichever model runs it.

And a high Landing score means a strong first draft, not a page you publish unread.

Members of the AI Profit Boardroom see these model picks set up inside the Agent OS, job by job.

Pick one job, pin one model, and compare the drafts for a week.

Then bring your results to the weekly calls in the AI Profit Boardroom and compare notes with other members.

Also On Our Network

• Agent Operatorsthe eight-step operator setup for a draft-only Hermes marketing desk
• AI Income Desksix small-business marketing use cases and the risks an owner should weigh
• AI Tool Verdictan eight-criteria scorecard rating Hermes on research, drafts and send safety
• agentos.guidefive copy-paste prompts that turn Hermes into a draft-first marketing desk
• aiprofitboardroom.comthe complete Hermes Agent for Marketing guide on the AI Profit Boardroom blog

Pin the right model to your first landing-page job this week, and book a free strategy session if you want help picking models for Hermes Agent for marketing.