Shorter answers did not fully offset GPT-5.5's higher pricing in OpenRouter's analysis of users switching between OpenAI's GPT-5.4 and GPT-5.5. For prompts above 10,000 tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →, median completions were 19–34% shorter. Yet the observed cost per million OpenRouter-counted tokens rose 49–92% across prompt-size groups. The analysis compares text requests from the same user cohort over short windows in April 2026 before and after launch. It measures that cohort's usage, rather than replaying identical tasks, so the percentages describe observed costs rather than a guaranteed increase for every application.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
