OpenAI API Pricing History

OpenAI's pricing history reflects a strategic shift toward model specialization. Recent updates include the introduction of the GPT-6 Astra family and various GPT-5.6 derivatives like Luna, Sol, and Cyber. These models are priced according to context length, speed, and modality, moving away from a one-size-fits-all token rate.

For procurement and finance leaders, the landscape now includes complex variables such as batch discounts, audio-specific token rates, and seat-based licensing for ChatGPT teams. Understanding these historical shifts is critical for forecasting the total cost of ownership as organizations scale from pilot programs to full-scale AI adoption.

OpenAI API pricing history

NetLift has monitored the official OpenAI API pricing page since 2026-09-09. 48 change(s) recorded so far.

Current verified prices

Current verified per-token prices

How does the GPT-6 Astra structure impact unit costs?

GPT-6 Astra introduces significant price variance based on context and speed. Standard short-context inputs are priced at $10 per 1M tokens, while long-context inputs double to $20. Output costs for these models are significantly higher, reaching $75 per 1M tokens for long-context tasks. Organizations can mitigate these costs by using Batch mode, which reduces short-context input costs to $5 per 1M tokens.

What is the difference between ChatGPT seat billing tiers?

Subscription pricing is now bifurcated by seat type and billing frequency. Standard seats cost $20 monthly on an annual contract or $25 on a month-to-month basis. Premium seats, designed for higher-tier access, require a larger investment of $100 per month (annual) or $125 (monthly). Both tiers require a minimum of two users, while non-profit organizations may access discounts of up to 75%.

Determining whether these API costs are justifiable requires moving beyond simple token counting to value-based measurement. By using NetLift, finance teams can baseline the time saved on specific workflows against the varying costs of models like GPT-6 Astra or GPT-5.6 Sol. This ensures that the premium paid for high-context or fast-mode tokens is generating a verifiable net return on the investment.

Sources

All pricing on this page comes from official vendor pages:

• https://platform.openai.com/docs/pricing — retrieved 2026-09-09
• https://openai.com/business/pricing/ — retrieved 2026-09-15