← Back to search

OpenCode Cat Jam

AI News Today | Julian Goldie Podcast · 2026-10-16 · 6 min
relevance 86 1475 words Episode page ↗ Audio ↗
Show full episode description
OpenCode Cat Jam
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Slow API-backed agents and lost build artifacts, addressed with OpenCode's free stealth model Cat Jam inside an agentic operating system.
Benefits
  • Very fast agentic coding, roughly twice as fast as Big Pickle
  • Free with unlimited usage for a limited time
  • Zero data retention; user data not used for training
  • Swap between OpenCode's other free models when rate limited
  • Workspace live previews keep every built artifact findable
Use cases
  • Built a to-do list app in about 10 seconds with Cat Jam
  • Adam Roman gave Cat Jam and Big Pickle the same prompt; working result in half the time
  • Plugging Cat Jam into Hermes Agent as a fast model for sub-agents and custom workflows
  • Hermes Oracle tracks news for content ideas; Hermes Muse analyzes past content for new ideas
KPIs / results
  • To-do app generated in ~10 seconds
  • Cat Jam ~2x faster than Big Pickle
  • Data retained up to 30 days for safety, not training
  • Dropped the day before recording
Tools / build
  • OpenCode with Cat Jam stealth model
  • Agentic operating system with workspace live preview
  • Hermes Agent (Hermes Oracle, Hermes Muse, pop-up voice agent)
  • Free fallback models: Big Pickle, Longcat 2.5, Step 5 Preview, MIMO V216, Muse Spark
  • AI Profit Boardroom community
0:00 / 0:00
There's a brand new model inside OpenCode that you can get access to called Cat Jam. It's a new stealth model and it's designed to be number one, very fast, number two, built for agentic coding, and number three, zero data retention, which I guess is pretty good for privacy and that sort of thing. So this is pretty wild. If you actually have a look here, Adam Roman, shout out to Adam, actually gave Cat Jam the same prompt as Big Pickle. We got something working in half the time. They actually preferred Big Pickle, which is another free model from OpenCode on the UI, but Cat Jam is like twice as fast as Big Pickle for building, which is more than twice as fast, which is pretty crazy when you look at the results right there. So basically, if you're wondering who this comes from, nobody knows, right? It's a new free model, it's stealth, but it's designed to be very fast and it'll also be really good inside something like Hermes Agent because it can respond so quickly. I would say that's one of the biggest issues that I see with Hermes Agent. It's like if you plug in an API, sometimes it can be a little bit slow. Whereas with this system, we can use OpenCode inside our agentic operating system, which we've got over here. And then we can go over to OpenCode on the left hand side and we can just start using Cat Jam for free. Now there's loads of free models on OpenCode, so that's pretty cool as well. So if you get rate limited, you can just switch between them. And then, for example, we can switch to Cat Jam, which is the free new fast model. And we say just, you know, basically, are you working? And then it should start working right here as well. The one thing I was thinking was like, if it's got zero data retention, does that mean the data doesn't go to servers? Or does that mean like it literally just can't remember anything? And there's not much information on Cat Jam because it's a brand new model. I mean, it only just dropped yesterday, as you can see right here. But still, looks pretty cool, pretty interesting. If you're wondering, you know, what it is and stuff like that, basically, it's a new and honest stealth model available for free on OpenCode. The key details, it's free for a limited time, but with unlimited usage, apparently. Performance built and optimized specifically for agentic coding, fast generation speeds, data privacy, so zero data retention from the provider and user data is not used for model training, which is unusual for stealth models, by the way. Like normally a stealth model, the reason that it's available for free, that it's using the data that it gets to improve it. But this isn't using the data from the provider. So if you start building with it, it's designed to not be used for model training, which is good to know. However, what it does say is retained for up to 30 days normally, depending on the provider policy. And then, of course, the exact creator and underlying model remain unconfirmed. But some people are saying it might be like Quen or Gemini variants. So it's really difficult to find information on Cat Jam. But basically, it's a free model available for a limited time. It's provided it does not use your data for model training, retains it for about 30 days for safety reasons. So that's the only reason it retained it, but it's not used for training. It's also not available in certain countries, as you can see right here. But for the majority of countries, it is available. And then also, if you're like, oh, do you know what? I'm in one of those countries and I still want to use it, but I can't. What I'd recommend is you just use one of the other free models, right? So, for example, Longcat 2.5, I've actually personally tested out Step 5 Preview. Really, really good. Really fast model. It's not like Opus 5.5 level, but it's pretty good and it's free to use as well. So if we go back to OpenCode right now, you can see it replies really quickly. If we say, for example, okay, build out a beautiful to-do list app. Just for fun. Let's test this out. And then the great thing about this is when we're using this inside our agentic operating system, we can have it working on writing and wow, that was insanely fast. What on earth? That was 10 seconds to build a to-do list app, which I've never seen before. I remember back in the days when I first started using Bolt with local models and it would take like 30 minutes to run and then slow everything down. So that's pretty crazy. And then also what happens is this goes into our workspace here. So we can have a live preview of everything we've built. So, for example, you can see we created this cool little fun thing with Cat Jam, as you can see. We've got the to-do list over here. We can open up any of these. The great thing about this is like if you're using this inside an Argentic operating system, it's way easier to keep track of everything you've built with. So you can see everything you've created and you can come back to it anytime. Whereas, for example, if you're using OpenCode normally, which is still awesome, you know, especially that you get free models. Like there's nothing wrong with that. But if you actually go over to OpenCode like you can see, it's like where's all the stuff you've created? How do you find previous stuff that you created last month? How do you visually preview everything if you can't even find it in the first place? You can't. You literally can't, right? So instead, what you want is something like you can see right here, which is just much easier and much more valuable, I think, to use. And then, for example, we can go back to the terminal and we can start building out with our AI agents over here. And you can see that it just works really nicely. Now, we could also plug this in as a model inside Hermes Agent. And then you can use it inside custom workflows. So, for example, we have a pop-up voice agent over here. We have Hermes Oracle that tracks the latest news and gives us a breakdown of new content ideas. We've got Hermes Muse that's just constantly analyzing my content and then giving me new ideas for new content based on what just works previously. So all this sort of stuff is really cool. And you can just plug this into your agentic workflows like you can see right here. So pretty cool model. Crazy fast. I think it's nowhere near Opus 5.5 level. So for anyone watching, they're like, oh, should I use Cat Jam? No, that's just dropped. No, I don't recommend that. But if you need a fast agentic model, if you need to build quickly, if you need a quick sub-agent and you just want to use a free model, this is great. And then if you do get rate limited, just switch to one of the other free models like MIMO V216 or, for example, Muse Spark. You've got Longcat. All these other. You've got Space Money. I don't know if that's still free, but that's on the list. And then you can switch between them and use them whenever you want. So thanks so much for watching. If you want to get more training on open code, how to use it, if you want to get the full agentic operating system that I've just shown you, if you want all my best trainings on AI agents and everything else, feel free to get that inside the AI Profit Boardroom. Link in the comments description or just go to the AI Profit Boardroom.com. And inside this community, you can post questions, get help and support. And it's just a great community to meet people. Loads of people using open code. Loads of people using Hermes Agent inside there. Inside the classroom, you can get access to all my best trainings like you can see. In the calendar, you can jump a weekly coaching calls, ask questions, get help and support in real time. And then in the map, you can meet people in your local area who are building out with AI agents like you. So feel free to get it. Link in the comments description or just go to the AI Profit Boardroom.com. Thanks for watching. Thanks for watching.