GPT-5.5 Price Increase, What's the Real Cost?

OpenRouter compared actual costs by selecting a cohort of users who switched from using GPT-5.4 as their primary model to GPT-5.5. Both models use the same tokenizer family, so tokenizer differences were not separately controlled for, and as OpenAI stated, 5.5 tended to produce shorter completions. Input tokens rose 2x from $2.50/M → $5.00/M, and output tokens rose 2x from $15/M → $30/M.

Response length varied depending on prompt size. At 10K tokens and above, GPT-5.5 was 19–34% shorter, but in the 2K–10K range, it was actually 52% longer. Below 2K, the length of the two models was nearly identical.

Based on OpenRouter's own token counting, the average converted cost rose across all ranges.

• < 2K: $4.89 → $9.37, +92%
• 2K–10K: $2.25 → $3.81, +69%
• 10K–25K: $1.42 → $2.15, +51%
• 25K–50K: $1.02 → $1.65, +62%
• 50K–128K: $0.74 → $1.10, +49%
• 128K+: $0.71 → $1.31, +85%

The analysis included only text-only, non-canceled requests, using data from April 21–23, 2026 for GPT-5.4 and April 25–28, 2026 for GPT-5.5. Image, file, audio, and video requests, as well as 0-token requests, were excluded, and the GPT-5.5 launch date was also excluded.