OpenAI has announced limited testing of its latest flagship model, GPT-6 Astra, which offers unprecedented computing power.
Nvidia CEO Jensen Huang said that more than 100,000 servers based on Nvidia Grace Blackwell chips were used to train the system, and the company plans to expand the infrastructure to 400,000 graphics accelerators.
The announcement came with a loud statement: Huang directly announced the realization of the era of AGI (artificial general intelligence), although OpenAI management urged us to treat these terms with caution and recognize the vagueness of the concept. The model itself has demonstrated record-breaking results on complex benchmarks such as FrontierMath Tier 4 and ARC-AGI 3, and the campaign focuses on advanced voice interfaces and versatility. However, this announcement comes amid the recent automated agent security scandal and rumors of an imminent initial public offering (IPO) of the company, which is why experts are calling for a new phase in the AI arms race to be entered with considerable skepticism.