01What the Beam Hermes Agent model is and how it works

The Beam Hermes Agent model is a brand new model that you can use for free for one week inside Hermes Agent. Julian Goldie shows what it is and how it works in this episode.

Beam is free in the portal for one week. It comes from Reflection, which is a US AI company. It is designed to be very fast and agentic when you use it.

Agentic means the model is built to act like an agent and get work done, not just chat. That is why it fits so well with Hermes Agent.

  • Free: no cost for one week in the portal.
  • Fast: built to reply quickly.
  • Agentic: made to power agents.
  • Open weights: you can use it across many different setups.

Julian says you can plug Beam into any of the agents you already run. He has already plugged it into his agentic operating system.

02Main features: fast, agentic and open weights

Beam has a few clear features that Julian points out in the episode. Each one matters if you build agents for real work.

  • Speed. Beam replies quickly, so chats and agent jobs do not feel slow.
  • Agentic design. It is made to be a powerful system for agents, not only for simple questions.
  • Open weights. Beam is an open weights model, so you can use it across any setup you want.
  • A US maker. Reflection is a US AI company.

The last point matters to some people. Julian explains that some people do not like to use Chinese models such as DeepSeek Flash. Step five is another Chinese model.

If that is you, he says fair enough. You can switch to Reflection and Beam instead. Julian also notes that Reflection has already served 1 trillion tokens of Beam in 24 hours, which he calls pretty wild.

03How to get started with the Beam Hermes Agent model

Julian shows a simple way to start using Beam. He uses a separate profile so Beam stays apart from his other models.

  1. Open Hermes Agent.
  2. Set up a new, separate profile for Beam.
  3. Pick Beam, which is free in the portal for one week.
  4. Start chatting with your agent to test it.
  5. Plug the profile into your own workflows or agent setup.

Julian has already done this inside his agentic operating system. In Hermes Agent he has a profile just for Beam.

The same screen lets him switch between other free models. This means you can try Beam, then change to another model in seconds. You do not have to commit to one model for all your work.

Remember that the free window is one week, so plan to test it soon.

04Switching between Beam and other free models

One of the best parts of the setup is that Beam is not your only choice. Julian shows that you can move between free models inside Hermes Agent.

For example, he also has DeepSeek 1017. This is a brand new model that dropped at the weekend. It is also free to use.

It is free because of the brain of OpenCode. OpenCode is the free way to get access to DeepSeek 1017. Julian then plugged that into Hermes Agent as well.

  • Beam: free for one week from Reflection.
  • DeepSeek 1017: free through the brain of OpenCode.

This gives you two free options in one place. You can test both on the same task and keep the one that suits your work. Julian says Beam works across any setup, so changing models does not break your flow.

05Testing the speed of the Beam Hermes Agent model

Julian tests Beam with a simple chat. He types a casual message to his agent, asking how it is going, and speaks to it like a normal person.

As he shows on screen, the reply comes back pretty fast. He says it is really quick and works pretty nicely. He also notes that Beam has a good little piece of personality in its answers.

This test is small, but it shows the two things Julian likes most. Beam is quick, and it feels natural to talk to.

Julian also says Beam is actually pretty useful, not only fast. He is clear about the limit, though. Beam is not at Opus 5.5 level, and you should be 100% aware of that.

It is not frontier level.

At the same time, Beam is designed to be agentic and to be a powerful system. That is the job it is built for.

06Where you can use Beam: Hermes tools and workflows

Julian shows that Beam works in many places across his setup. You are not stuck with one screen.

  • Hermes chat for normal conversations with your agent.
  • Hermes bot screen for running agent work.
  • Hermes Apollo as another agent you can pair it with.
  • Hermes Oracle for another kind of agent job.
  • Hermes Muse to analyze your existing content.

Hermes Muse is a powerful way to see how your content is performing. It then gives you recommendations for content to make in the future.

You can also use Beam for SEO machines or whatever else you want. Julian also says you can plug it into your own custom agentic workflow. Because Beam is open weights, it can sit inside any of these different setups.

07Memory, limits and who the Beam Hermes Agent model is for

Most people are not using Beam day to day, so how do you train it up and sync it with memory? Julian answers with a shared memory system.

His system holds all of his memories and projects. His AI agents update that memory every time he uses them. When he speaks to Hermes Apollo or uses Hermes Astros, it is all added to the memory vault.

This means a new model like Beam can work with the same memory as your other agents.

  • Use it if you want a free model inside Hermes for one week.
  • Use it if you want a fast, agentic, open weights model.
  • Use it if you prefer a US maker over Chinese models.

Remember the limit. It is not frontier level. Julian also says his full agentic operating system makes Hermes and every other agent 10 times more powerful.