The economics of AI are starting to favor open models

For the last couple of years, the assumption was pretty simple: Want the smartest model? Pay for a closed API. Want something cheaper? Accept a capability hit. Looking at recent model releases, that tradeoff is starting to break down. The most interesting part of the chart isn't the models at the very top. It's the upper-left quadrant. High intelligence. Low cost. And it's increasingly dominated by open-weight models. DeepSeek. Qwen. GLM. Kimi. MiniMax. Most real-world workloads don't need the absolute best model on Earth. They need a model that's: Good enough Cheap enough And that's exactly where open models are becoming incredibly competitive. A year ago I would've assumed the gap would stay huge because the frontier labs had access to significantly more compute and data. For a lot of tasks, the difference between a frontier model and a strong open model is becoming smaller than the difference in cost. That's a dangerous trend if you're selling expensive API tokens(and good news for everyone else lol) Closed models still have advantages: Zero infrastructure Better reliability Faster access to frontier capabilities But open models offer something APIs never can: (i mean some do say things like trust me bro im secure and give full privacy but u cant take them on their word) Full control Privacy Customization Predictable costs My prediction: Within 12-18 months, most businesses won't be asking: What's the smartest model? They'll be asking: Why am I paying 10x more for a 5% improvement? and how does it compare to the open source stuff