OpenAI has two things today, and seeing them together is especially interesting:
First: GPT-5.6 Sol is half off on OpenRouter and Vercel when they hit their limits, while the official API price stays unchanged. Second: GPT-6—its code name was leaked again. It’s said to be released as early as Thursday: a native multi-agent architecture, with a $2,000 token cost to solve 10 tenth-grade math problems.
SemiAnalysis immediately starts criticizing: “Half off? Your share in total usage across OpenRouter and Vercel is tiny, but the market’s AI market share estimates mainly rely on those two metrics—you’re basically ‘manipulating’ the market-share numbers.”
Well, you’re discounting old models to push the metrics, while the new model is arriving right away. That move is kind of like a milk tea shop: second cup half off to clear inventory, then new products launch right after?
I’ll ask just one thing: if GPT-6 really is coming this Thursday, will your API budget be enough? $OPENAI
First: GPT-5.6 Sol is half off on OpenRouter and Vercel when they hit their limits, while the official API price stays unchanged. Second: GPT-6—its code name was leaked again. It’s said to be released as early as Thursday: a native multi-agent architecture, with a $2,000 token cost to solve 10 tenth-grade math problems.
SemiAnalysis immediately starts criticizing: “Half off? Your share in total usage across OpenRouter and Vercel is tiny, but the market’s AI market share estimates mainly rely on those two metrics—you’re basically ‘manipulating’ the market-share numbers.”
Well, you’re discounting old models to push the metrics, while the new model is arriving right away. That move is kind of like a milk tea shop: second cup half off to clear inventory, then new products launch right after?
I’ll ask just one thing: if GPT-6 really is coming this Thursday, will your API budget be enough? $OPENAI