Breaking

OpenAI Releases GPT-6 Astra, Reigniting Debate Over Jensen Huang’s AGI Claim

September 6, 2026 at 17:44 EDT

Foundation Models
AI Agents
Security

OpenAI has released GPT-6 Astra, a frontier model built for long, tool-using work, while the launch has renewed scrutiny of Jensen Huang’s claim that artificial general intelligence has already arrived. The chronology matters: Huang made that statement months before Astra’s release, so the available record does not show that it was a response to this model.

September 3, 2026 · OpenAI

GPT-6 Astra Meets the AGI Question

OpenAI’s new frontier model targets long, tool-using work. Its benchmark gains and safety disclosures sharpen the debate over what AGI means in practice.

98%FrontierMath Tier 4

99.9%ARC-AGI-3

100%ExploitBench

1.9xMind2Web task speed

ExploitBench: reported cyber capability

Higher scores are better. Each block represents 5 percentage points.

100%

GPT-6 Astra

78.5%

GPT-5.6 Sol

The AGI question

Capability is measurable. AGI remains a definition.

Jensen Huang’s March claim used a threshold of an AI system capable of starting, growing, and running a company worth more than $1B. GPT-6 Astra’s release does not establish that threshold.

What the launch enables

Longer computer-use workflows, coding, research, and polished documents.

What remains unresolved

Staged access, higher cyber risk, and lower chain-of-thought monitorability.

Continue reading

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