OpenAI launches GPT-6 Astra with computer-use tools across ChatGPT, API, Azure and Bedrock

OpenAI has launched GPT-6 Astra, a model it positions as its most capable and aligned to date, built around advanced computer use, software engineering, browsing, cybersecurity tasks and professional knowledge work. According to a report on dev.to, the release is a staged, multi-platform deployment: an initial phase for a limited group of organizations, followed by availability for paid ChatGPT users and developers across the OpenAI API, Microsoft Azure and AWS Bedrock.

What the model does

The dev.to report frames Astra's defining capability as working across tasks that normally require moving between applications, web pages and business documents. OpenAI positions the model for filling in forms, updating CRM records, managing calendars, researching the web, installing and troubleshooting software, and producing documents, spreadsheets and presentations that follow a user's templates and style.

That is a wider remit than a language model used to draft text or answer questions. OpenAI also points to strong performance in code generation and professional knowledge work, with the stated ability to install, test and troubleshoot software pointing toward more autonomous technical workflows.

How access works

The rollout is staged rather than simultaneous. Astra reaches a limited set of organizations first, then expands to ChatGPT Plus, Pro, Business and Enterprise users over the following days. Third-party reporting cited by dev.to describes an initial "Daybreak-program" release ahead of the broader expansion.

Because access may differ by account, workspace configuration and platform during early deployment, the report notes that workspace administrators may need to enable the model for Pro, Business and Enterprise workspaces, and that developers should confirm API access, quotas and pricing configuration before committing Astra to a production workflow. Availability through OpenAI, Azure and Bedrock does not guarantee identical implementation conditions; eligibility and quotas should be checked in each provider's documentation.

Specs and pricing

The published limits and access paths:

• Context window: up to 1,050,000 tokens.
• API pricing: $10 per million input tokens and $50 per million output tokens.
• Maximum output: up to 128,000 output tokens.
• API model name: gpt-6-astra.
• Fast mode: an option that doubles speed at a higher price.
• ChatGPT tiers: usage included within existing allowances, with additional credits available for purchase.

A context window of that size matters when a workflow needs to consider extensive source material — large document collections, codebases or operational records — in a single request. The report is careful to note that it does not by itself make a process accurate or suitable for automation; teams still need to define the task, permissions, review points and acceptable outcomes.

The $10/$50 per-million-token rates give a clear starting point for cost estimates, but the key operational distinction is between occasional assistance and an automated process that repeatedly reads large inputs or produces long outputs. The latter can consume tokens quickly, especially when tasks lean on Astra's large context capacity.

Safety claims

OpenAI says Astra reaches a Critical level of cybersecurity capability under its Preparedness Framework. The company describes stronger protections around the model, including stricter isolation, enhanced chain-of-thought monitoring, misalignment monitoring, improved adversarial robustness and safeguards intended to keep the model within authorized scope.

The report also cites zero-data-retention options for API customers, which may matter for teams assessing where model interactions fit within data-handling requirements — though it does not remove the need to review what data a workflow sends to a model and what actions it is allowed to take.

Why it matters

A model that can navigate authorized tools and complete multi-step digital tasks targets the repetitive administrative work that sits between systems in most organizations: copying information between tools, assembling routine client documents, researching defined sets of sources or running repeatable software tasks. If the computer-use functions prove reliable, that is a genuine expansion of what API customers and ChatGPT subscribers can automate.

The counterweights are cost and control. Greater task autonomy raises the value of clear permissions, limited initial deployments and human review for consequential actions, particularly where workflows touch customer records, calendars or other operational systems. OpenAI's safeguards are part of the product's stated posture, not a substitute for careful workflow design. For any given team, Astra's value will come down to whether the task is stable enough to automate, whether the model has authorized access to the right systems, and whether the time saved outweighs token and implementation costs.