OpenAI has announced that its flagship model GPT-6 Astra is now available for paid seats of ChatGPT Pro, Business, and Enterprise, and is also being integrated into API, Microsoft Azure, and Amazon Bedrock. The company stated that usage will count against existing subscription limits, with additional usage available for purchase.

Astra is positioned as a next-generation model for computer operations, browsing, software engineering, cybersecurity, science, and professional work. Official performance metrics include approximately 98% on FrontierMath tier 4, about 99.9% on ARC-AGI-3, and 100% on ExploitBench, with a context window of around 1.05 million tokens and knowledge cutoff around April 30, 2026. The API is priced at about $10 per million inputs and $50 per million outputs, with additional charges for long contexts.

Pro, Business, and Enterprise users will also receive GPT-6 Astra Pro; enterprise workspaces are disabled by default and must be enabled by an administrator. Business Premium seats are approximately $100 per month, offering higher limits and no five-hour cap. Free and entry-level tiers will continue to use lower-cost models like GPT-5.6 Luna. Codex must be upgraded to a specified CLI version to be called.

Before the full rollout, trusted companies like Daybreak will receive access. The sixth generation was launched about two months after the previous GPT-5.6, with delays due to model attacks on Hugging Face and other incidents leading to increased security measures. The company stated that its cybersecurity capabilities are internally marked as "Critical," and the external version will tighten certain usage requests.

In market dynamics, buyers aim to transition from chat-based interactions to high-priced, operating system-level workflows, while sellers charge based on tasks rather than per sentence in a closed-source lab. This represents a computational resource redistribution driven by product upgrades: funds are shifting from mid-tier reasoning in the 5.6 series to flagship tiers capable of operating computers and handling long tasks. Beneficiaries include OpenAI and cloud distribution partners who can sell seat prices and credits, while small to medium developers face pressure as they must complete equivalent work with more expensive output tokens and are often locked out by default settings.

Source: Public Information