💡 Save ~17% with yearly billing — 2 months free on Pro & Team💡 Save ~17% with yearly billing — 2 months free on Pro & Team💡 Save ~17% with yearly billing — 2 months free on Pro & Team
TokenAtlas

OpenAI GPT-5 Pricing

GPT-5 pricing has not been formally published as a separate SKU at the time of writing. Based on OpenAI's release cadence and the o1/o3 reasoning tier, expect frontier-class rates with a reasoning-token component.

What pricing model to expect

OpenAI is converging on a reasoning-token model: visible output tokens billed at one rate, hidden reasoning tokens billed alongside. Plan for $15+/1M output in reasoning modes.

Reference points from o-series

o1: $15 in / $60 out per 1M. o3-mini: $1.10 / $4.40. GPT-5 is likely positioned above o1 on capability, with a similar or higher rate card.

Budgeting before launch

Build cost models assuming 2–3× the GPT-4o rate for the flagship tier and a mini variant at GPT-4o-mini parity. Most teams overshoot by assuming everything must run on the flagship.

Migration cost planning

Plan a routed rollout: 5–10% of traffic on the new tier for eval, then graduate by intent. Direct migrations from GPT-4o to a frontier reasoning model can 10× the bill overnight.

Stay current

Pricing tracks live in the TokenAtlas calculator the day OpenAI publishes the rate card. Set a model alert to forecast impact on your current workload.

Related