The Headline Numbers
OpenAI's GPT-5.5 lands at $5.00 per million input tokens and $30.00 per million output tokens — a sharp step up from GPT-5.2's $1.25/$10. For teams that need the absolute ceiling, GPT-5.5 Pro is available at $30/$180. The mid-range is covered by GPT-5.4 at $2.50/$15, with GPT-5.4 Nano at $0.20/$1.25 for high-volume simple tasks.
| Model | Input $/1M | Cached Input | Output $/1M |
|---|---|---|---|
| GPT-5.5 Pro | $30.00 | — | $180.00 |
| GPT-5.5 | $5.00 | $0.50 | $30.00 |
| GPT-5.4 | $2.50 | $0.25 | $15.00 |
| GPT-5.4 Nano | $0.20 | $0.02 | $1.25 |
Why the Cache Discount Is the Real Story
The 90% cached-input discount is the number that matters for production systems. Agent loops, RAG pipelines, and chat applications re-send the same system prompt, tool definitions, and conversation history on every request. If 80% of your input tokens hit the cache — common for agents — your effective input rate drops from $5.00 to about $1.40 per million tokens. Suddenly GPT-5.5 is competitive with models half its sticker price.
GPT-5.5 vs the Field
Against Claude Opus 4.8 ($5/$25), GPT-5.5 charges the same for input and 20% more for output. Against Claude Fable 5 ($10/$50), it is roughly 40% cheaper across the board. Gemini 3.1 Pro undercuts it at $2/$12, and Grok 4.3 ($1.25/$2.50) costs a tenth as much on output — the question, as always, is whether your workload notices the quality difference. Batch API processing remains 50% off for non-urgent workloads.
Should You Upgrade From GPT-5.2?
If you run a reasoning-heavy workload — agentic coding, multi-step analysis, complex tool use — the upgrade likely pays for itself in fewer failed runs. If GPT-5.2 was already saturating your task (most chat and content generation), the 4x list-price jump buys you little. Test before you migrate.
Run your own numbers in our AI cost calculator, or see the full head-to-head in Claude Fable 5 vs GPT-5.5.