OpenAI's latest flagship model GPT-5.5 carries a significant price premium over its predecessor, with costs rising between 49 and 92 percent depending on how long the input is, according to an analysis by API aggregator OpenRouter. The company doubled the list price for GPT-5.5 compared to GPT-5.4, arguing that the model's ability to generate shorter responses would balance out the higher per-token costs. However, OpenRouter's examination of actual usage patterns reveals that users are not experiencing those savings—the real-world cost increase consistently falls in that 49 to 92 percent range across different input lengths.
The pricing analysis suggests that while GPT-5.5 may produce more concise outputs in some scenarios, the savings do not compensate for the steeper token prices. OpenRouter's data indicates that for shorter inputs, the cost increase hovers near 49 percent, while longer inputs push that increase toward 92 percent. This discrepancy between OpenAI's marketing narrative and observable user costs highlights a gap between theoretical efficiency claims and practical billing outcomes.
Anthropic has similarly raised prices for its Claude Opus 4.7 model, following a broader industry trend among AI companies seeking to improve margins on expensive compute infrastructure. Both OpenAI and Anthropic are reportedly preparing for initial public offerings, which creates strong pressure to demonstrate revenue growth and profitability. Industry observers expect this upward pricing trajectory to continue as these companies work toward meeting investor expectations ahead of potential IPOs.