It all started with this tweet that came up on my feed.
OpenRouter.ai – I haven’t used OpenRouter much, but I did have an account there. So I immediately went to it to try this out. What’s interesting about OpenRouter is that, in some cases, it lets you chat completely for free with AI models.
This particular model, Quasar Alpha, which is on par with—if not better than—some of the best foundational models I’ve tested, is currently in stealth mode. We don’t know which company created it. Still, it’s making waves in AI circles on X thanks to its generous 1M token input window and a reasonable 35k token max output window.
comparison of quasar alpha vs sonnet 3.7 thinking
The model is cloaked to avoid revealing its origin, but you can interact with it for free on OpenRouter.ai.
Most remarkable of all is how fast it generates high-quality outputs—on par with, if not better than, leading foundational models like Gemini 2.5 and Claude 3.7. I haven’t done any extensive testing yet, but I encourage you to try it as soon as possible.