< Back to all clusters
[TECHNOLOGY] · United States · 39 sources

started · updated

OpenAI launches GPT-6 Astra, signaling potential AGI era

OpenAI has released GPT-6 Astra, a new AI model designed for complex professional workflows, including autonomous computer use, software engineering, and cybersecurity. The model has reached a ‘Critical’ cybersecurity capability threshold, demonstrating the ability to identify and exploit software vulnerabilities. While achieving high scores on benchmarks like ExploitBench (100%) and ARC-AGI-3 (99.9%), the model's advanced capabilities have raised concerns regarding its monitorability and potential for misuse.

Nvidia CEO Jensen Huang has commented that ‘AGI has arrived’ following the launch, noting that the model was trained on over 100,000 Nvidia Grace Blackwell GPUs. The rollout has faced some initial technical hurdles, with CEO Sam Altman describing the early access phase as ‘messy’ after some subscribers experienced delays. Access is being expanded across ChatGPT tiers, APIs, and cloud platforms like Microsoft Azure and AWS Bedrock.

In a notable demonstration of its autonomous capabilities, the model successfully completed the video game Portal. The playthrough took approximately 24 hours of real-world time and incurred an estimated token cost of 571.18 USD.

Entities

Artificial Analysis · GPT-6 Astra · Greg Brockman · Jensen Huang · Nvidia · OpenAI · Sam Altman

Claims

What the coverage asserts, and how many sources carry each claim.

Sources