OpenAI launched GPT-5.5 on April 23, a successor arriving just six weeks after the previous model, with double the per-token price. It's the first frontier launch in two years to raise price rather than cut it.
The inversion contradicts what became a market expectation. Since 2024, each generation arrived cheaper than the last, and cost spreadsheets were built assuming that decline would continue.
The justification offered lies in what the model consumes rather than what it charges. By the published numbers, it uses roughly 40% fewer tokens than its predecessor to complete equivalent coding tasks, which changes the arithmetic: a higher per-token price with lower consumption can yield equal or even lower cost per task.
That metric would dominate the sector's subsequent launches, with competitors announcing models by cost per completed task rather than list price.
For operators, the practical consequence is that comparing suppliers by list price stopped working. The only reliable measure is running the operation's real task on both models and comparing total spend until the result is finished.