OpenAI launches GPT-6 Astra with staged plan and API access

GPT-6 Astra is rolling out as OpenAI’s new top-end model for complex reasoning, coding, computer use, research and document creation. The launch begins with enterprises in OpenAI’s Trusted Access Program; Plus, Pro, Business, Enterprise and API access are scheduled to follow in the coming days.

GPT-6 Astra targets end-to-end knowledge work

OpenAI positions Astra for tasks that span multiple steps and tools rather than a single answer. The official model page names reasoning, software development, computer interaction, research and document creation as its target workloads. That positioning makes the release relevant to teams already building workflows around long-running agentic tasks.

The model supports five reasoning-effort levels: low, medium, high, xhigh and max. Developers can therefore trade response depth against latency and cost within the same model family. OpenAI says Astra is usable through the Responses API, with support for the tools listed on its model documentation page.

Access expands in stages across OpenAI plans

The current lifecycle stage matters. GPT-6 Astra is available today to enterprises in the Trusted Access Program, while the broader Plus, Pro, Business and Enterprise rollout is described as coming in the next few days. The API is also part of that staged expansion. OpenAI’s wording does not establish universal availability or a precise activation timestamp for every account.

For teams planning an integration, the practical checkpoint is the model catalog and the model guide. An account seeing the model page should still verify API eligibility, regional availability, rate limits and the applicable pricing before moving production traffic.

Pricing rises sharply with long-context and fast modes

OpenAI describes pricing as usage-based and directs developers to its pricing page for the final rates. Requests above 272K input tokens are priced at twice the input and cache rates, with output priced at 1.5 times the standard output rate for the full request. Cache writes cost 1.25 times the uncached input rate; Batch and Flex use half the Standard rates, while Fast mode doubles the applicable rates.

Artificial Analysis independently lists GPT-6 Astra (max) at $10 per million input tokens and $50 per million output tokens, with a 1M-token context window and support for text and image input. Those figures should be treated as a current external measurement of the max variant and checked against OpenAI’s live pricing page before budgeting.

What developers should verify first

The release creates a clear evaluation path for teams with demanding coding or research workloads:

• Confirm whether the intended OpenAI plan or API account has received access.
• Test each reasoning-effort setting on representative tasks instead of assuming max is always best.
• Measure the cost impact of requests above 272K tokens and of Fast mode.
• Check tool compatibility in the Responses API before migrating an existing agent.

For workflow design ideas, LinkLoot’s AI workflow automation guide is a useful companion. The next concrete milestone is broader account and API availability, followed by independent results from real production workloads.