Do Son 
According to updated OpenAI support documentation, the GPT-6 Astra context window now supports up to one million tokens. Furthermore, users utilizing GPT-6 Astra within Codex no longer face surplus charges for content exceeding 272,000 tokens.
Policy Application and Testing
This revised billing policy applies exclusively to the GPT-6 Astra model functioning inside Codex. OpenAI continues to apply surplus charges for exceptionally long content across other distinct models. Users should always conduct preliminary tests to verify billing before committing to extensive long-term deployment.
Previous Premium Pricing Models
Earlier models also supported one-million-token contexts. However, OpenAI previously imposed premium fees for extended content. The standard rate covered the initial 272,000 tokens. Any usage beyond that strict limit incurred multipliers ranging from 1.5 to 2 times the standard fee. Consequently, users exhausted their Codex quotas rapidly when configuring extended context support.
Updated Billing Rate Structures
For the GPT-5.4, 5.5, and 5.6 series, standard rates apply up to 272,000 tokens. Beyond this specific threshold, OpenAI bills input tokens at double the standard rate. They also charge double for cached inputs, while output tokens incur a 1.5x multiplier. OpenAI outlines these detailed tier structures in their enterprise token-based pricing rate card.
For GPT-6 Astra operating within Codex, input exceeding 272,000 tokens no longer triggers multiplier rates. Additionally, Codex completely eliminates charges for cached inputs. Therefore, all operations within the entire one-million-token window now utilize the standard baseline rate.
Support Our Threat Intelligence
Find our zero-day alerts and CVE reports helpful? Support our work today and unlock a 100% ad-free reading experience!
Buy Me a Coffee
PayPal
