OpenAI’s GPT-6 Astra Launches Big, but Paying Users Are Left Waiting

OpenAI calls GPT-6 Astra a leap toward an AGI era, with powerful computer-use and cyber capabilities. But an enterprise-first launch triggered subscriber frustration, forcing Sam Altman into a public apology and a promise of broader access.
OpenAI’s GPT-6 Astra Launches Big, but Paying Users Are Left Waiting

OpenAI’s GPT-6 Astra Launches Big, but Paying Users Are Left Waiting
OpenAI unveiled GPT-6 Astra on Thursday as its most powerful model yet, pitching a system that can operate a computer, tackle professional work and strengthen cyber defenses. The company’s public message was unmistakably ambitious: Altman said Astra could enable “a new generation of entrepreneurship, scientific discovery, and building.”

The technical case was equally bold. Greg Brockman said the computer-use capability could “zip through spreadsheets, fill out forms, and navigate across web pages often at superhuman speed.” OpenAI’s reported benchmark results put Astra well ahead of its previous GPT-5.6 Sol model on ARC-AGI-3 and at 100% on ExploitBench, while Perplexity chief executive Aravind Srinivas called it the industry’s frontier model and said it was “far ahead” on broad research tasks.

But the rollout quickly exposed the gap between the launch rhetoric and the customer experience. Astra initially went to a limited group of enterprise clients, including customers using OpenAI’s Daybreak cybersecurity platform. The broader Plus, Pro, Business and Enterprise release was promised for the following days, leaving many paying ChatGPT subscribers — especially Pro users accustomed to early access — locked out.

The restricted rollout also reflected a safety calculation, not simply capacity constraints. Astra is OpenAI’s first model to meet its internal threshold for critical cybersecurity capability, meaning it can find and exploit unknown vulnerabilities under certain conditions. Daybreak partners can use it for approved defense work, but the consumer-facing version is intended to refuse advanced cyber tasks.

Within hours, Altman conceded the launch had gone badly: “first, sorry for the messy rollout.” He said broad access for API customers and ChatGPT subscribers should begin “in the near future,” starting with Pro accounts; the quoted company response also promised one banked reset for each day paid users lacked access. The episode leaves OpenAI trying to sell Astra as a democratizing breakthrough while rationing its most consequential capabilities — and, initially, access itself.

Continue reading https://foxvector.com/stories/01a06d11-b2a2-355a-7195-0970b12ad64a

Write a comment