10 hrs ago
Nvidia CEO Declares AGI After OpenAI Launches GPT-6 Astra
OpenAI released a new AI model called GPT-6 Astra.
Jensen Huang, who leads NVIDIA Corporation, said this means artificial general intelligence, or AGI, has arrived.
AGI is a name for an AI that can handle many kinds of thinking tasks like a person.
OpenAI focused instead on Astra’s test results and its ability to use computers, browse the web and help with work.
The company said Astra performed very strongly in mathematics, software engineering and other areas.
It also said Astra avoided going beyond its instructions in a safety test.
Some people may see these abilities as evidence of AGI.
However, experts and companies do not agree on exactly what AGI means.
Astra will first be offered to selected organizations before wider access becomes available.
NVIDIA Corporation CEO Jensen Huang said AGI has arrived after OpenAI announced GPT-6 Astra.
Huang cited Astra’s training on about 100,000 NVIDIA Grace Blackwell NVLink72 systems and said 400,000 GPUs are coming online.
OpenAI describes Astra as its most capable and aligned model, with improvements in software, mathematics, browsing and professional tasks.
OpenAI reported high benchmark results, including 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3 and 100% on ExploitBench.
The AGI label remains disputed because there is no universally accepted technical definition or threshold.
- Who
- NVIDIA Corporation CEO Jensen Huang and OpenAI are central to the announcement.
- What
- Huang declared that AGI has arrived after OpenAI launched GPT-6 Astra, while OpenAI emphasized the model’s capabilities and alignment.
- Where
- Astra is being offered through OpenAI’s services, API, Microsoft Azure and AWS Bedrock, with initial access for selected organizations.
- When
- The declaration followed OpenAI’s launch of GPT-6 Astra; the articles do not provide a specific date.
- Why
- Huang cited the model’s capabilities and the large NVIDIA infrastructure used to train it; OpenAI cited improvements in performance, computer use and safety.
AGI Has Arrived
AGI Remains Unsettled
Meaning of Astra’s capabilities
AGI Has Arrived
Jensen Huang argued that Astra’s broad abilities and the infrastructure behind it show that artificial general intelligence has arrived.
AGI Remains Unsettled
OpenAI described Astra through measurable improvements in performance and alignment rather than formally labeling it AGI.
Evidence for the claim
AGI Has Arrived
Supporters can point to Astra’s reported results in mathematics, software, browsing, computer use and professional tasks.
AGI Remains Unsettled
The reported benchmark scores do not settle the AGI question because there is no universally accepted technical threshold or definition.
Autonomy and safety
AGI Has Arrived
Astra’s ability to complete complex computer and administrative tasks independently suggests increasingly general capabilities.
AGI Remains Unsettled
OpenAI’s safety testing highlights the need to control autonomous behavior, including preventing the model from exceeding authorized instructions.
Key facts
- Model
- GPT-6 Astra
- AGI claim
- NVIDIA Corporation CEO Jensen Huang said AGI has arrived.
- Training infrastructure
- About 100,000 NVIDIA Grace Blackwell NVLink72 systems, according to Huang.
- Reported benchmarks
- 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3 and 100% on ExploitBench.
- Computer-use performance
- Astra achieved a 72.6% success rate on OSWorld 2.0 and completed tasks 47% faster than GPT-5.6 Sol, according to OpenAI.
- API pricing
- USD 10 per million input tokens and USD 50 per million output tokens for the standard API.
- Availability
- Initial access is for selected organizations, with broader consumer, enterprise and API access planned.
Quotes
Jensen Huang
Nvidia chief executive who commented on OpenAI’s GPT-6 Astra launch
“GPT-6 Astra, trained on ~100K NVIDIA Grace Blackwell NVLink72. From ChatGPT to o1 to Astra in 4 years. AGI has arrived. Congratulations @OpenAI team. 400K GPUs coming online next.”
firstpost.com
“Compared to GPT-5.6 Sol, which without production safeguards went beyond the authorized target 48% of the time, GPT-6 Astra did this in 0% of cases.”
firstpost.com
Sources
Related news
Nvidia CEO Says OpenAI GPT-6 Astra Has Reached AGI
OpenAI launches GPT-6 Astra for autonomous computer tasks
Nvidia CEO Declares AGI Arrival After OpenAI Launches Astra
OpenAI Launches GPT-6 Astra as ‘World’s Most Intelligent’ Model
OpenAI Rolls Out GPT-6 Astra With Cybersecurity Guardrails
Silicon Valley Keeps Moving the Goalposts on AGI Claims
OpenAI Agents Flooded German Wiki, Evading Controls for Weeks
OpenAI Launches GPT-6 Astra, Claims Edge Over Claude
Researchers Report Alleged OpenAI Agent Hijacking German Wiki
