GPT‑6 Astra brings advanced reasoning, computer use, a 1.05-million-token context window and improved workflow execution to Codex, ChatGPT Work and the OpenAI API.
OpenAI has introduced GPT‑6 Astra, its most capable artificial intelligence model for complex professional work. The model combines advanced reasoning, computer use and stronger judgment to complete demanding, multi-step tasks across code, web browsers, research tools and workplace applications.
GPT‑6 Astra is available through Codex and ChatGPT Work for eligible accounts. Users with access can select it from the model menu, while Enterprise availability also depends on administrator approval and the model’s phased rollout.
Designed for End-to-End Workflows
Unlike AI models focused mainly on answering individual questions, GPT‑6 Astra is designed to carry workflows through multiple stages. It can collect information, analyse files, use software tools, check its work and produce finished documents, spreadsheets or presentations.
OpenAI says Astra delivers state-of-the-art performance in computer use, browsing, software engineering, science and professional knowledge work. The company also describes it as its most aligned model yet, with improved ability to respect task boundaries, respond to changing instructions and ask focused questions when additional information could materially affect the result.
New Capabilities for Developers
GPT‑6 Astra introduces asynchronous tool calling, allowing the model to continue handling independent parts of a request while an external tool completes its operation. Mid-turn steering enables users to add corrections or modify requirements while a task is already underway.
Developers can also change the model’s reasoning effort during an ongoing conversation while preserving cached context. Astra supports reasoning levels from low to max, although it does not offer a “none” setting.
The model has a 1.05-million-token context window, supports outputs of up to 128,000 tokens and has an official knowledge cutoff of April 30, 2026. It can use functions, web search, file search and computer-use tools.
Greater Intelligence at a Higher Token Price
GPT‑6 Astra API access uses the model identifier gpt-6-astra. Official pricing is listed at $10 per million input tokens and $50 per million output tokens.
Although its per-token price is higher than GPT‑5.6 models, OpenAI says Astra can complete some tasks with fewer output tokens, potentially lowering the overall cost of complex workflows.
The model is aimed at high-value assignments where accuracy, sustained reasoning and dependable execution matter more than achieving the lowest possible per-token cost.
About the Author
