OPENAI’S GPT-6 Astra within Codex now supports a context window of up to one million tokens, with the notable change that users are no longer charged surplus for content exceeding 272,000 tokens. The article clarifies that this revised billing policy applies only to the GPT-6 Astra model when used inside Codex; other models remain subject to previous long-context charges.
In earlier pricing, OpenAI applied multipliers for content beyond 272,000 tokens (ranging from 1.5× to 2× the standard rate) and charged separately for cached inputs, while output tokens carried a 1.5× multiplier. The update removes these excess charges for Codex users, meaning operations within the full one-million-token window now accrue only the standard baseline rate.
The piece notes that the changes are specific to Codex usage and contrasts this with ongoing surplus charges for exceptionally long content on other model configurations. It also highlights a relevant CVE in its CVE Watchtower feed: CVE-2026-85046, described as a type confusion issue in Google Chrome’s V8 engine prior to 152.0.7977.82, which had prompted active exploitation signals.
Practically, organisations deploying GPT-6 Astra within Codex are advised to verify billing during testing and to recognise that the one-million-token window is billed at the standard rate, with no added penalties for cached inputs. The article stresses confirming pricing details before long-term deployment.