Breaking

OpenAI rolls out GPT-6 Astra with frontier computer-use and cybersecurity gains

September 3, 2026 at 22:27 EDT

Foundation Models
AI Agents
Security

OpenAI introduced GPT-6 Astra on September 3, positioning it as a model for complex, end-to-end work across computer use, coding, research and professional workflows. The release combines stronger benchmark results with unusually strict controls around its cybersecurity capabilities.

September 3, 2026 · OpenAI

GPT-6 Astra moves AI agents into complex work

OpenAI says Astra combines advanced reasoning with computer use, coding and research, while its strongest cyber capabilities remain tightly controlled.

99.9%

ARC-AGI-3 score

100.0%

ExploitBench score

1.05M

Context window

Cyber capability rose, and so did the guardrails

Same benchmark, same percentage scale: each block represents 10 percentage points.

78.5%

GPT-5.6 Sol

100.0%

GPT-6 Astra

How Astra approaches open-ended work

Define the goal

→

Build a plan

→

Use approved tools

→

Deliver for review

What users value

Complex workflows can be completed with fewer handoffs, clearer artifacts and stronger computer interaction.

What remains hard

Advanced cyber work is restricted, monitoring is still difficult and safeguards may interrupt legitimate tasks.

Capability and control now ship together

Astra expands access over the coming days, with permissions, monitoring and human checkpoints shaping how far its agents can act.

Continue reading

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