OpenAI rolls out GPT-6 Astra to ChatGPT Plus subscribers
The new model introduces 'agentic' capabilities and high-benchmark performance for paid users.
OpenAI has begun the gradual rollout of GPT-6 Astra, its most powerful model to date, to ChatGPT Plus subscribers. The release marks a strategic shift toward AI agents capable of executing complex, end-to-end professional tasks.
Access is currently expanding to users on the Plus ($20/month), Pro, Business, and Enterprise tiers. According to OpenAI, the model is rolling out to a limited set of organizations first before becoming available to all paid subscribers over the coming days. Users may first see the model appearing in the 'ChatGPT Work' and 'Codex' sections before it integrates into the standard chat interface. Beyond the consumer app, Astra is also available via the OpenAI API, Microsoft Azure, and AWS Bedrock. While paid tiers are receiving access, OpenAI has not yet provided a timeline for free users.
A Generational Leap in Intelligence
Released on September 3, 2026, as a limited preview for trusted partners, GPT-6 Astra follows previous iterations such as GPT-4o and GPT-5.6 Sol. The model was pre-trained on more than 100,000 GPUs at OpenAI's Stargate site in Texas, representing a massive scale-up in compute resources.
Technically, Astra is designed to move beyond simple conversational responses. It focuses on "agentic" capabilities, including the ability to browse the web, use a computer, and perform professional software engineering. Astra is significantly better than its predecessor, GPT-5.6 Sol, at maintaining context during complex work, which is critical for long-form professional projects.
The Path Toward AGI
This release signals a pivot toward AI that can actively operate a computer similarly to a human. By automating high-level workflows in fields such as cybersecurity, science, and coding, Astra could fundamentally alter professional productivity.
The model's performance on specialized benchmarks suggests a significant advance toward Artificial General Intelligence (AGI). Astra achieved a 98% score on FrontierMath and a 99.9% score on ARC-AGI-3, figures that indicate a level of reasoning and mathematical capability far exceeding previous LLM generations.
What to Watch
As the rollout continues, the industry will be watching to see how Astra handles real-world computer use and whether its high benchmark scores translate into reliable professional autonomy. While the technical capabilities are confirmed, the broader impact on the labor market for software engineers and data scientists remains the primary point of speculation. OpenAI has yet to clarify if the model's advanced agentic features will be gated behind higher-priced tiers or available to all Plus subscribers.