GPT-5.4 API Pricing & Cost Calculator

How this is calculated

GPT-5.4 is priced at $2.50 per million input tokens and $15.00 per million output tokens, supporting a 90% caching discount ($0.25 per million) and a 50% batch discount ($1.25 per million inputs, $7.50 per million outputs).

Verdict

At $2.50/M input and $15.00/M output a representative 100K + 5K call costs about $0.325, half GPT-5.5's $0.65 per call. A 90% cache hit strips input to $0.25/M, and the 50% batch discount drops both rows to $1.25/M input and $7.50/M output. That puts GPT-5.4 in the rare spot of carrying frontier-tier reasoning across the bulk of production traffic without wrecking the bill. Reserving it for cache-heavy agentic loops and routing stateless calls elsewhere keeps the per-token cost low enough to scale.