OpenAI Introduces GPT-6 Astra With Advanced Computer Use And Coding Capabilities

OpenAI has introduced GPT-6 Astra, a new flagship AI model designed to complete complex tasks involving software development, research, computer use, and document creation. The company describes Astra as its most capable model. OpenAI’s model documentation

The release follows the GPT-5.6 family, which introduced Sol, Terra, and Luna on July 9, 2026. Astra arrives less than two months later, adding another model to OpenAI’s lineup for professional and technical work. OpenAI’s API changelog

Access is beginning with enterprises participating in OpenAI’s Trusted Access Program. The company says availability through its API and ChatGPT Plus, Pro, Business, and Enterprise plans will follow in the coming days. OpenAI’s model catalog

GPT-6 Astra Targets Multistep Computer And Browser Tasks

OpenAI says Astra can carry out workflows spanning code, browsers, and professional software. Its stated areas of strength include browsing, software engineering, science, and other knowledge-based work.

The model also introduces features that let it continue reasoning or address independent parts of a request while an external tool is running. Users can provide corrections or new instructions during supported ongoing sessions.

According to OpenAI, Astra is designed to incorporate those changes without losing track of the broader assignment. The company also reports improvements in following task boundaries and communicating transparently. OpenAI’s Astra guidance

OpenAI Adds Monitoring For Sensitive AI Actions

Astra’s safeguards include monitoring intended to identify situations in which an agent may be misinterpreting instructions during consequential tasks.

The system reviews model reasoning and actions involving activities such as accessing sensitive information, transferring data, or making destructive changes. Depending on the API configuration, it can issue alerts or stop a conversation.

OpenAI cautions that the monitoring can miss problems or flag legitimate activity. It recommends maintaining separate application safeguards, including human approval for consequential actions. Stopping a conversation also does not reverse actions already completed. OpenAI’s monitoring documentation

GPT-6 Astra API Pricing And ChatGPT Availability

Standard API pricing for Astra is $10 per million input tokens and $50 per million output tokens. Cached input costs $1 per million tokens.

Those rates increase for requests exceeding 272,000 input tokens. OpenAI lists long-context pricing at $20 per million input tokens and $75 per million output tokens. Batch and Flex processing offer lower rates, while Fast mode costs more. OpenAI’s pricing page

OpenAI argues that token prices alone do not capture the cost of completing a task. The company says Astra used fewer output tokens in several evaluations, producing a lower estimated API cost per task than earlier models despite its higher token rates. OpenAI’s Astra guidance

For ChatGPT subscribers, access remains subject to the phased rollout rather than immediate availability across every eligible account.