# GPT-6 Astra

GPT-6 Astra is OpenAI's flagship model announced on September 3, 2026, described as the product of years of research; it is the first OpenAI model to reach the company's Critical internal cybersecurity threshold and is rolling out in phases.

GPT-6 Astra is a flagship [large language model](https://www.wikiprompt.org/wiki/large-language-model) developed by [OpenAI](https://www.wikiprompt.org/wiki/openai), announced on September 3, 2026. The company described it as the product of "years of research and big bets" and began a phased rollout to ChatGPT Plus, Pro, Business and Enterprise plans, the OpenAI API and Amazon Web Services.

## Capabilities and safety gating

Astra is the first OpenAI model the company disclosed as reaching its "Critical" internal cybersecurity capability threshold under its preparedness framework, a category of [responsible scaling commitments](https://www.wikiprompt.org/wiki/anthropic-principles) that triggers access restrictions before release. OpenAI said the most advanced cyber-relevant capabilities would be limited: the first group to receive access is a set of companies in Daybreak, OpenAI's application-based cybersecurity program. The gated launch mirrors the two-track pattern [Anthropic](https://www.wikiprompt.org/wiki/anthropic) uses for its [Fable and Mythos](https://www.wikiprompt.org/wiki/claude-fable-5) tiers, where the most capable variant is restricted to vetted organizations.

## Context

The launch closed an unusually dense week for [frontier models](https://www.wikiprompt.org/wiki/frontier-model): Anthropic released Claude Fable 5.1 and Mythos 5.1 on September 1, and Google shipped Gemini 3.8 Flash on September 2. Astra follows OpenAI's GPT-5.6 family (July 2026), which introduced the Sol, Terra and Luna variants, and continues the industry shift toward [reasoning-first](https://www.wikiprompt.org/wiki/reasoning-model) flagships evaluated as much on safety thresholds as on capability [benchmarks](https://www.wikiprompt.org/wiki/benchmark).

## Reception

As of its launch day, GPT-6 Astra had not yet appeared on public community leaderboards such as [LMArena](https://www.wikiprompt.org/wiki/lmarena); phased rollouts typically reach the arenas within days. Coverage focused on the cybersecurity gating as a milestone: it was the first time OpenAI publicly held back parts of a flagship's capability surface at launch under its own preparedness thresholds, a practice safety researchers had urged since the [Bletchley commitments](https://www.wikiprompt.org/wiki/bletchley-declaration) and one likely to shape how future [governance frameworks](https://www.wikiprompt.org/wiki/ai-governance) treat dual-use capability disclosures.

---
Source: https://www.wikiprompt.org/wiki/gpt-6-astra
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-03T19:12:33.416267+00:00