GPT-5.5 Price Increase: What It Actually Costs
Analyze token usage per prompt size and adjust billing thresholds to mitigate cost spikes.
Calculate token usage per prompt size and adjust billing thresholds to mitigate cost spikes.
Summary
GPT‑5.5 launched with a 2× price increase over GPT‑5.4: input tokens cost $5.00/M instead of $2.50/M and output tokens cost $30/M instead of $15/M. The model is less verbose, producing shorter completions for the same tasks. Analysis shows cost increases between 49‑92% across prompt sizes, mitigated by 19‑34% fewer completion tokens for longer prompts. The study used a switcher cohort of users who moved from GPT‑5.4 to GPT‑5.5, keeping the tokenizer constant for a direct before‑and‑after comparison. Completion length changes by prompt size are: <2K +7%, 2K‑10K +52%, 10K‑25K –32%, 25K‑50K –19%, 50K‑128K –28%, 128K+ –34%. Cost per million tokens rose 49‑92% across all buckets, with <2K +92%, 2K‑10K +69%, 10K‑25K +51%, 25K‑50K +62%, 50K‑128K +49%, 128K+ +85%. Longer prompts over 10K offset cost by shorter completions, but short prompts still see a higher cost increase.
Key changes
- GPT‑5.5 launched with input token cost $5.00/M and output token cost $30/M, double GPT‑5.4 rates
- Completion length increased by +52% for 2K‑10K prompts but decreased by 19‑34% for prompts over 10K tokens
- Average cost per million tokens rose 49‑92% across all prompt size buckets
- The model produces 19‑34% fewer completion tokens for prompts over 10K, partially offsetting higher input costs
- The analysis used a switcher cohort of users who moved from GPT‑5.4 to GPT‑5.5, keeping tokenizer constant
- Completion length changes: <2K +7%, 2K‑10K +52%, 10K‑25K –32%, 25K‑50K –19%, 50K‑128K –28%, 128K+ –34%
- Cost per million token changes: <2K +92%, 2K‑10K +69%, 10K‑25K +51%, 25K‑50K +62%, 50K‑128K +49%, 128K+ +85%