The Throne is Never Secure

@阑夕:If you only look at the "Big Three" of American large language models, the past year truly feels like a rotating cast taking turns in the spotlight, with each leading the pack for a few months:

Around this time last year, ChatGPT's text-to-image feature ignited the market. The "Ghibli style" created an epic viral sensation, directly doubling the weekly active users of a platform that was already far ahead of its competitors;

Then came the second half of the year, the "Google's home turf," where Nano Banana fought back with its multimodal capabilities. Gemini 3 firmly established its SOTA (State-of-the-Art) status, and Google's comeback was highly celebrated by the media;

Next, from the end of last year to the beginning of this year, Claude swept through the software industry. Anthropic, entirely focused on the B2B market, achieved the miracle of rapidly boosting its ARR to number one globally purely through model capabilities;

And now? It seems OpenAI has woken up again. GPT-Image-2 staged a king's return in the text-to-image domain, GPT-5.5 Instant didn't flop, and Codex's reputation and metrics are even subtly overshadowing Claude Code;

The throne is never secure.