OpenAI's GPT-6 Now Comes in Two Versions: One Cheap, One Brilliant

OpenAI has introduced two new AI models, GPT-6 Sol and Luna. Rather than one giant brain that does everything, the company is offering a pair: one tuned for raw capability, the other tuned to be cheaper to run. Both are aimed at ordinary work tasks — writing, summarizing, answering questions, handling documents — rather than research stunts.

Why does a split like this matter to you? Because the biggest hidden cost of AI at work isn't the software fee, it's how much computing power each question burns. Every time an AI reads a long email thread or answers a customer, someone pays for that. If a cheaper model can handle the boring 80% of tasks, companies can afford to give AI to every employee instead of a handful. That's the difference between AI being a perk for your manager and a tool on your own desk.

The trade-off is real, and it's the part to watch. Cheaper models tend to make more mistakes — they may miss a detail in a contract, misunderstand a sarcastic email, or confidently invent a fact. So the practical question for any business isn't "which one is smarter," it's "which one is smart enough for this specific job, cheaply?" Expect a lot of trial and error, and expect some embarrassing early mistakes at companies that get the switching wrong.

The bigger picture: AI models are becoming a product line, not a single product. Just as carmakers sell economy and luxury versions of the same vehicle, AI companies are now competing on price as much as power. For you, that likely means AI gets cheaper and more common in your daily tools over the next year — and that your employer will increasingly expect you to use it. One honest catch: OpenAI hasn't said what Sol and Luna cost, how well they perform, or when you can try them, so the real-world verdict is still unknown.