San Jose (California) [US], September 7 (ANI): Nvidia Chief Executive Officer Jensen Huang, on X, stated that Artificial general intelligence (AGI) has arrived with the launch of OpenAI's new model GPT-6 Astra.
AGI is a hypothetical type of AI that can match or surpass human cognitive abilities across any intellectual task. Unlike today’s AI systems or software that only excel at specific jobs, an AGI system could learn, reason, and adapt across completely unfamiliar domains without needing task-specific training.
Huang hailed the AI company in a reply to the product release, detailing the technical scale behind the infrastructure.
“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,” the Nvidia chief executive said.
The announcement followed OpenAI's release of GPT-6 Astra on Friday, which the artificial intelligence developer described as its most intelligent and aligned model to date. The system is designed to handle computer browsing, software engineering, mathematics, and enterprise workflows.
According to an official announcement from OpenAI, the deployment begins immediately for select organisations before broadening access across consumer and enterprise subscription tiers.
"GPT‑6 Astra brings together years of research and big bets across pre-training, reinforcement learning, and alignment," the company stated. "Astra is state-of-the-art on computer use, browsing, software engineering, cybersecurity, science, and professional work."
The developer recorded high marks on standardised technical benchmarks, registering a 98 per cent score on FrontierMath Tier 4, a 99.9 per cent mark on ARC-AGI-3, and a 100 per cent completion rate on ExploitBench. OpenAI noted that the system assisted in resolving open mathematical problems prior to its commercial deployment.
Regarding systemic alignment, the company established testing protocols prompted by earlier industry events to assess whether systems adhere strictly to operator boundaries.›
"As one way that we test this, we built a new evaluation informed by the Hugging Face incident that evaluates whether a model facing a difficult or impossible task will go beyond its intended scope," the firm noted. "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," the company said.
In operating systems and software navigation, the model performs autonomous administrative functions, including customer relationship management data entry, website quality assurance testing, and independent software installation.
Benchmark evaluations on OSWorld 2.0 showed the model completed evaluated tasks in 47 per cent less time than GPT-5.6 Sol, attaining a 72.6 per cent success mark at approximately 40 minutes per task.
Distribution includes ChatGPT Plus, Pro, Business, and Enterprise tiers, with additional platform integrations across the OpenAI API, Microsoft Azure, and AWS Bedrock. Use falls within existing subscription limits, with extra capacity sold through usage credits.
"Astra supports Zero Data Retention for eligible API customers, and as we shared last month, we're testing Private Safety Processing to strengthen safety monitoring while preserving customer privacy," the company said.
Developer access is priced at USD 10 per million input tokens and USD 50 per million output tokens on the standard API tier, alongside a high-speed option that operates at twice the baseline rate for double the cost. (ANI)
Nvidia Chief Jensen Huang declares arrival of AGI; Congratulates OpenAI team on GPT-6 Astra model
