Breaking

OpenAI begins limited rollout of GPT-6 Astra ahead of wider access

September 3, 2026 at 21:23 EDT

Foundation Models
AI Agents
Security

OpenAI has begun rolling out GPT-6 Astra to a limited group of enterprises, while wider access through paid ChatGPT plans and the API is scheduled for the coming days. The company’s most advanced cybersecurity workflows remain gated to a small group of testers after Astra became its first model classified at the Critical level under its Preparedness Framework. OpenAI’s rollout guidance

September 3, 2026 · OpenAI

GPT-6 Astra starts with restricted access

OpenAI has begun a limited rollout of GPT-6 Astra, with broader paid-plan and API access planned in the coming days. Its most advanced cybersecurity workflows remain limited to testers.

Rollout at a glance

1.05M

Context tokens

128K

Maximum output

100%

ExploitBench score

$10/$50

Input / output per 1M

Reported workflow result

Manual fixes fell by half

100%

Previous model

50%

GPT-6 Astra

Playco reported 50% fewer manual fixes in its game-prototyping workflow.

Developer signal

Works inside real tools

Playco used Astra with Unity and Godot to build, test and refine playable prototypes.

Safety limitation

Advanced cyber use is gated

Astra reached OpenAI’s Critical cybersecurity threshold and can be less monitorable under adversarial testing.

The rollout is phased by design

Enterprise access begins first; paid plans and API access follow, while advanced cybersecurity workflows remain limited to testers and defensive-access programs.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.