The model becomes outdated on the day of release: OpenAI released three updates in two days, and it took only five hours to catch up again.
## Models become outdated on the day of release: OpenAI and others announced three releases in two days, with the fastest overtaking only taking five hours On September 4th, OpenAI announced the phased release of GPT-6 Astra, opening it to all Plus users first. The official statement was that “stability and speed were both important,” and the entire release was expected to be completed within a few days. Just two days earlier, on September 2nd, Google and Meta each released new models on the same day. Google launched Gemini 3.8 Flash, its third Flash model in six weeks; Meta released Muse Spark 1.3, its fourth Spark model in five months. Hours after Meta’s new model went live, it overtook Google’s new products on the Artificial Analysis intelligence index. There were only forty-eight hours between the two releases. Two years ago, the release of a flagship model would have dominated the headlines for a week. Today, the shelf life of cutting-edge models has been compressed to hours. This is not a problem of just one company; it’s a more serious sign: large-scale model intelligence itself is becoming something that can…