OpenAI's GPT-5.5 Pricing Doubles, But Real User Costs Rise 49-92% Based on OpenRouter Analysis
Key Takeaways
- ▸OpenAI's GPT-5.5 carries a 2x official price increase: input tokens $2.50→$5.00/M, output tokens $15→$30/M
- ▸Real-world cost increases for switching users range from 49-92%, with variation based on prompt length and usage patterns
- ▸GPT-5.5's reduced verbosity (19-34% fewer tokens) only applies to longer prompts (>10K tokens); shorter prompts see minimal token reduction
Summary
OpenAI launched GPT-5.5 with a headline-grabbing 2x price increase over its predecessor, raising input token costs from $2.50/M to $5.00/M and output token costs from $15/M to $30/M. However, a detailed cost analysis by OpenRouter reveals the real-world impact on users is more complex. Researchers tracked "switcher cohort" users who migrated from GPT-5.4 to GPT-5.5, finding actual cost increases ranging from 49% to 92%—significantly higher than a naive calculation might suggest.
The key finding is that GPT-5.5 mitigates its price increase by being less verbose, but only for longer prompts. The model generates 19-34% fewer completion tokens for prompts exceeding 10,000 tokens, providing meaningful cost relief. However, for shorter prompts under 2,000 tokens, completion lengths remain roughly equivalent, meaning users with shorter workflows absorb nearly the full price increase. Users with prompts in the 2K-10K range see completions that are actually 52% longer, compounding costs in that segment.
The analysis, based on OpenRouter's consistent token counting methodology across both models, provides the most granular public breakdown of GPT-5.5's actual cost impact to date. It suggests the model's verbosity reduction—while real—is not evenly distributed and may not justify the price increase for all user segments.
- Users with shorter workflows and mid-length prompts (2K-10K tokens) see the steepest relative cost increases, as completion lengths remain unchanged or even increase
