Qwen3.6 Hands-On Test – Is THIS a Frontier Coding Model?

What is this? It signed this apology with humility and respect. Your AI assistant slows the frame rate momentarily. So you For today's video, we're going to be taking a look at the somewhat newly releasleased Quen 3.6 Plus model. Now, this was released really soon after the 3.5 model, and basically a lot of what is outlined here as major improvements relate to Agentic capabilities. So for today's test, we are going to be testing it, of course, in our traditional way, just through the web chat interface for some of the simpler tasks. However, additionally to that, we're going to be using it through open code as well, just to give it more ability to showcase its agentic capabilities. So, we're going to begin with a quick look at the announcement post here. Talk about how this is not currently open source, which I suppose is just talking about it then. Um, there's not much to say about that. And hopefully this will be released in the future as open source as well as the iterative updates that will follow for the smaller models in the 3.5 series. Beyond that, I do just want to take a moment to say that I am almost going to hit 50,000 subscribers. And I want to just take a second and say I should be way higher. This is ridiculous. But I'm kidding. Feel free to subscribe though. And it's cool to hit that milestone, I suppose could be said. So we'll jump into looking at this introductory announcement post and then we'll get to testing. So to begin, let's take a quick look at some of the interesting things in this announcement post where we can see the Quen Capabro, which is the mascot, has also been given a claw just in keeping with the very high popularity of OpenClaw. We begin by seeing some specific benchmark JPEGs. And also, we get to know about the context window, which is 1 million by default. Now, I don't in this when I did just skim through it prior to filming this see any specific mention of the model size, but I would go out on a limb here and guess it is likely the same size as the Quen 3.5 Plus model, which was also open sourced in a different variant, but it was a 397b model with 17 billion active parameters. So, I do believe it's fair to say that this is the same specific size of that model. Quen 3.6 is also natively multimodal.