You Don't Need to Pay for Claude Anymore transcript

Google just dropped Gemma 4, and I know what you're thinking. Oh, another AI model, who cares? Here's why you should care if you're running Open Claw, or Mess Agent, or any other AI agent. Gemma 4's 31 billion parameter model is literally ranked number three in all open models in the world right now. It runs on normal consumer hardware. It's fully compatible with most AI agents straight out of the box. Meaning your AI agent, the one that automates your business, runs your crons, handles your content, can now run completely free without any API bills.

No Anthropic subscription for the tasks that you don't need it for. I mean, look at this. It's literally number three, right behind Kimmy K2 and GLM. And just to show you this, I mean, just look at this. This is amazing. This is Gemma 4. And look at how small it is as a model compared to the top two, which are GLM and Kimmy K2. This is insane. It is a fraction of the size of the top two, yet it's quickly rose to the top.

It's literally dominating that section right here. And I mean, let's look at this. This model was literally built for AI agents. Advanced reasoning, capable of multi-step planning and deep logic. Agentic workflows, native support for function calling, and native system instructions enables you to build autonomous agents. Code generation, vision and audio, longer context, 140 plus languages. Even Nvidia just published the guide today showing how to run it with Open Claw on RTX GPUs.

Hugging Face explicitly listed Open Claw compatibility in their launch blog. Here's how I think about this. Not every task my agents do needs Claude Sonnet, or Haiku, or Opus. Not every task needs a $200 subscription. My cron jobs, things like simple health checks, daily briefings, monitoring tasks, these don't need the best model in the world. They need a fast, reliable tool calling model. That's Gemma 4. My content strategy, my trading decisions, my complex reasoning, that can stay on Claude.

The best model can go for the hardest tasks. Gemma 4 can handle everything else for free, running locally on your machine. is literally if 60% of all your agent tasks are routine things like monitoring, scheduling, simple data pulls, and you route those to Gemma 4 locally, you could cut your API bill significantly. I can see people cutting more than half of their bill. Okay, so how do you set this up? Step one, you can install Ollama.

You can go to Ollama, copy the one-line installation, paste it in your terminal, hit enter, and just like that, Ollama starts installing, and boom, just like that, install completed. Step two, you can paste this Ollama pull Gemma 4 26B or 31B, depending on the size of the local machine you have, what it can run, what it can be compatible with, but that's all you need to do. You hit enter again, and boom, just like that, it starts downloading that local model into your machine.

It's that simple. And then step three, in your Open Claw configuration, or or Mess Agent configuration, point specific agents or cron jobs at local Ollama endpoint instead of Anthropic. That's it. Your agent now has two brains. Claude for the hard stuff and Gemma 4 locally for everything else. Look, I've been paying $200 a month for my Anthropic subscription, and I don't regret it whatsoever. Claude is genuinely the best agentic model for complex reasoning, but I've also been running tasks on Sonnet or Haiku that honestly don't need to be on those models.

Things like daily briefings and simple health checks. Gemma 4 changes that equation. This is the model I've been waiting for to do the hybrid setup that I had in mind properly. If you're also not on a GPU-enabled machine, the hosted version is on Open Router for almost nothing. Either way, I don't think there's an excuse to not be testing this out today. Anyway, download Gemma 4 from Hugging Face or through Ollama. I'll include the links in the description.

And if you want more content like this, then subscribe and follow for more. I'll see you in the next one.