§ 01 · The Hook
The bait, then the rug-pull.
A tweet from Teknium lands in frame: DeepSeek V4 Flash is back on Nous Portal for free. In eight minutes, WorldofAI turns that announcement into a working demo -- browser automation, scheduled research, front-end generation -- all running at zero API cost on a model that outperforms 77 of the 87 it was tested against.
Where the time goes.
00:00 – 01:19
01 · Introduction
Hook on DeepSeek V4 Flash landing free in Nous Portal; overview of Hermes Agent capabilities and the value proposition of the combination.
01:19 – 03:03
02 · How To Setup
Install Hermes locally (Linux/Mac/Windows beta), create a free Nous Portal account, run hermes model, select Nous Portal provider, pick DeepSeek V4 Flash free tier.
03:03 – 04:01
03 · DeepSeek V4 Usage
Live demo of Hermes running with the free model; Artificial Analysis benchmark data shown on screen confirming #10 ranking and 121 tok/sec speed.
04:01 – 04:34
04 · My Benchmarks
Creator own benchmark: DeepSeek V4 Flash vs Claude Opus 4.7 on a SaaS landing page front-end generation task, showing competitive output at zero cost.
04:34 – 05:44
05 · Research Demo
Hermes runs a scheduled research agent: web search across multiple sources, summarize AI model releases in the last 24 hours, compare benchmarks, output a Markdown report.
05:44 – 06:31
06 · Output
Markdown report converted to HTML via follow-up prompt; result is a readable but imperfect blog-style page opened in Cursor.
06:31 – 07:01
07 · Tools/Features
Overview of Hermes 19+ tool sets: browser use, skills, scheduled tasks, /goals command, usage tracking dashboard in Nous Portal.
07:01 – 08:48
08 · Frontend Output + CTA
Honest assessment: clear bugs visible in generated front-end; treat it as scaffolder then refine with Opus. Channel CTAs and sign-off.
Visual structure at a glance.
Named ideas worth stealing.
00:42
model
Hermes Skill + Memory Loop
Hermes accumulates long-term memory and builds reusable skills over time -- the longer it runs, the more capable it becomes for that specific user context.
Steal for Any recurring workflow where context from prior sessions reduces rework
07:17
concept
Scaffold-Then-Refine
Use a fast free model to generate structure (HTML, Markdown, code outline), then hand off to a stronger model to fix bugs and polish.
Steal for Cost-conscious multi-model pipelines where quality and speed both matter
Lines you could clip.
00:34
"You're essentially getting access to extremely powerful autonomous AI operating environments at zero cost with this combination."
03:06
"I'm not saying this is a perfect god tier workflow, but the value proposition here is honestly kind of insane."
07:17
"You can use this as a scaffolder and it can get the job done to a point where you can just simply use another model maybe like Opus to refine certain outputs."
Things they pointed at.
04:09productClaude Opus 4.7
How they asked for the click.
07:30
next-video
"Make sure you go ahead and take a look at the Universe of AI, which is our second channel. Join the newsletter. Make sure you also join the Discord."
Stacked CTA: second channel, newsletter, Discord, Twitter, subscribe. Preceded by Super Thanks mention. High frequency but delivered quickly without hard sell.
Word for word.
HOOK opening / re-engagementCTA the pitch
00:00HOOKLooks like Hermes agent just received another massive update today where deep seek version four is now completely free to use inside news portal, which is honestly a huge game changer. This basically means you're getting near state of the art reasoning, coding, long context, and autonomous agent performance completely for free inside an open source AI agent harness.
00:23HOOKNow this is where if you are to combine Hermes persistent memory system, a multi agent orchestration, browser use, computer control, and self improving workflows. You're essentially getting access to extremely powerful autonomous AI operating environments at zero cost with this combination.
00:42HOOKAnd for those who are unfamiliar with Hermes agent, it's one of the most interesting open source AI agent projects right now. It's designed as a persistent autonomous system that continuously evolves over time. It's built by news research under the MIT license, and Hermes agent can practically run twenty four seven off of your own infrastructure while building long term memory, reusable skills, and even deeper understanding of your user itself as you use it more and more.
01:10HOOKIf you want the best AI tools, workflows, and drops before everyone else, join my free newsletter with the link in the description below which is completely free. Now to get started with this setup, obviously, you need to make sure that you have Hermes agent installed locally. Now one great thing is that it is now supported for Windows, but it is currently in beta testing, but you can still get the Hermes agent installed
01:33onto your Windows operating system. Once you have the agent installed, you can then head over to News portal and what you gotta do here is just make an account and you can do this completely for free and you wanna select the free tier to access the News portal to get access to all the free models like DeepSeek version four.
01:50Once you have an account, then you can head over to the portal and we're gonna keep this open in the meantime. After selecting the free tier, what you can then do is open up a command prompt and within this command prompt, what you gotta do is just simply type in Hermes and then space and then type in model and then click enter.
02:07This is where you have the ability to now configure whatever model you wanna use. Currently, I have the codex plan linked up to my Hermes agent, but you can select number one which is where you're gonna be able to use the news portal. This will prompt open the news portal, uh, agent sign in.
02:25Once you have connected it to your free tier, you can then access it directly within Hermes agent and you'll see right away the DeepSeek version for Flash is completely free. So you can select that model and you can just simply click number one into your command prompt and click enter and this will default the model to use the DeepSeek version for Flash completely for free within Hermes.
02:47Now all you gotta do is just prompt open the agent itself by typing in Hermes and you can see that it is now gonna use the DeepSeek version for Flash completely for free directly within Hermes agent, letting you use all of these features completely for free. And I'm not saying this is a perfect god tier workflow,
03:06but the value proposition here is honestly kind of insane. This model is quite underrated actually. It is obviously extremely fast and you're getting state of the art reasoning and agentic performance completely for free through Hermes.
03:19And technically speaking, this matters a lot because this model is not some tiny lightweight model. In terms of intelligence, artificial analysis is stating that it is ranked number 10 in terms of its overall performance in the index. It is extremely fast, number eight out of 87 different models.
03:38Obviously, pricing is free. And in terms of overall usage, you can see that it is listed at a 121 tokens for approximately one second which is extremely
03:49efficient and it is extremely fast. This is something that also supports a massive 1,000,000 token context window. It is pretty good at reasoning and coding.
03:59It's surprisingly good for autonomous workflows. Now in my own benchmark, I will be releasing fairly soon, this model is extremely fast while also being something that can excel at front end, agenetic tasks, as well as system simulation.
04:14I'm not saying it's the best, but the value that you're getting for a free model is pretty good with this harness like Hermes. And you can see that it does pretty decent in terms of getting the basic structure and output for your front end in comparison to something like Opus 4.7 which probably is spending a lot more for an output like this.
04:33And what's cool is you can even use it like a research agent for example, which can work incrementally based off the schedule that it is given. And you can see that it's able to use tools like web search completely for free, something that's built within the news research portal.
04:49Currently, can see that it is scouring multiple sources to get my research tasks done where it's extracting multiple contents on what is happening right now throughout the AI model race where it is I'm trying to find different model releases within the last twenty four hours. It is also focusing on summarizing the biggest updates and then comparing the benchmarks
05:10and then at the end generating a clean markdown report with all the different sources. And just like that, it has created a markdown report which I've just opened this up within cursor and you can see that it has all the different sources as well and talks about all of the different findings as well as comparing it to other benchmarks as well as a baseline.
05:30So that is actually pretty interesting and that was a research task that I had completed right directly within my computer using a free model. Now what I'm gonna say is make this into a good looking report in HTML. And just like that, we have the HTML,
05:47a blog post that talks about all of the summaries that it was able to compile and research upon. And you can see that this is a pretty decent looking front end that it had generated, and that is what I mean.
05:58It's not something that's horrible and it's something that can actually get the task done pretty quickly and give you a good condensed report on multiple topics that you tell to research upon. I'm not saying this is something that you should just use for research, you can use it for smart file organization, Excel or as an AI,
06:17uh, analyst for whatever spreadsheets that you have. You can have it used for browser use workflows and so many other components that can essentially automate a lot of things in your life completely for free. And remember, you can use all of these available tools with this model.
06:33There's 19 plus tool sets that are directly available within your Hermes agent like browser use using different skills as well as setting up different scheduled tasks as well as their new slash goals command which you can use with the DeepSeek version four. You also have a lot of these different available skills. And if you wanna track your usage with the model, you can do so within the news portal and this way you can track your usage for your daily credit,
06:59your deep sea credit usage. This is what I mean by front end generation. It's not the best but it can get the job done and it can ease off a lot of your tasks when using this model for different use cases.
07:12CTAYou can see that there's a lot of clear bugs and a lot of things that would need a lot of refinement, but you can use this as a scaffolder and it can get the job done at to a point where you can just simply use another model maybe like Opus to refine certain outputs to make sure everything is thoroughly completed. If you like this video and would love to support the channel, you can consider donating to my channel through the super thanks option below, or you can consider joining our private discord where you can access multiple subscriptions to different AI tools for free on a monthly basis,
07:46CTAplus daily AI news and exclusive content, plus a lot more. Overall, this is a remarkable setup that lets you use Hermes agent completely for free with a pretty great model and this way you can do almost anything
08:00CTAwith this harness. Now this is something I would highly make the most out of within the next few weeks as this might change within the next few weeks. We don't know if they might revert back to a pricing tier for the DeepSeq version four.
08:13CTASo just keep that in mind, but I would highly recommend making the most out of this new update. I'll leave all the links that I used in today's video in the description below. But with that thought, guys, thank you guys so much for watching.
08:23CTAHope you enjoyed today's video. Make sure you go ahead and take a look at the universe of AI, which is our second channel.
08:29CTAJoin the newsletter to get more daily AI news. Make sure you also join the Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on notification bell, like this video, and please take a look at our previous videos so that you can stay up to date with the latest AI news. But with that thought, guys, have an amazing day.
— full transcript
§ 05 · For Joe
A free model with memory beats a paid model without it.
WHAT TO LEARN
The unlock is not the zero-cost model tier -- it is what persistent memory and a 19-tool harness do to the value of that model over time.
- DeepSeek V4 Flash ranks #10 on the Artificial Analysis intelligence index and #8 for raw speed across 87 models -- the free label does not mean underpowered.
- A 1,000,000 token context window means large codebases, long documents, and multi-source research tasks all fit in a single agent session without truncation.
- Persistent memory compounds: the longer Hermes runs, the more prior context it carries, making each subsequent task faster and more accurate for your specific workflows.
- The scaffold-then-refine pattern -- free model for structure, stronger model for polish -- reduces cost without sacrificing output quality on tasks that need both speed and precision.
- Hermes connects to Nous Portal with one command, making the switch from a paid provider to the free tier a single CLI interaction with no code changes.
- The free tier is time-limited by design; building workflow habits around the harness now means those habits persist even if the model pricing changes.
