[GPT-6 Astra Removes Long Context Surcharge in Codex: 1M Window at Standard Rate] Beating AI Newsflash: OpenAI has updated its pricing rules, granting a special exception for GPT-6 Astra within Codex. Astra inherently supports a 1.05 million token context, and even if the input exceeds 272K in Codex, it will no longer trigger the long context surcharge. Other models do not receive this benefit. When GPT-5.4, GPT-5.5, and GPT-5.6 use ultra-long contexts, once the input exceeds 272K, input and cached input are charged at 2x the rate, while output is charged at 1.5x the rate. Astra, however, can utilize nearly 1M context in Codex while still being billed at the standard token rate. Note that this discount applies only to Codex. The API will continue to enforce the 2x input and 1.5x output surcharge for long contexts. [Original Link]