OpenAI publishes GPT-6 Astra model that it is or is close to achieving AGI

On Thursday, 4 September, OpenAI began to publish the new flagship model GPT-6 Astra, stating that it had made significant progress in the areas of financial modelling, engineering and other commercial tasks. The model was initially open only to selected institutions, including organizations involved in the OpenAI Daybreak Access Network Security Project, and will then be open to users of ChatGPT Plus, Pro, Business and Enterprise, as well as users of the API and Amazon cloud services. OpenAI states that GPT-6 Astra achieved significant improvements in computer operations, software programming, professional work, etc., and reached the highest level in the Frontier Math Tier 1 mathematical test, ARC-AGI 3 reasoning test and Exploit Bench network security test. The co-founder and Managing Director of OpenAI, Greg Blockman, stated that there was always ambiguity in the timing of the implementation of General Artificial Intelligence (AGI), but that “a few years later, this milestone “may well be in this period, near this model”. GPT-6 Astra is priced at US$ 10 per million input and US$ 50 per output, 2.5 times the price of the previous flagship model GPT-5.6 Sol. OpenAI indicated that, in the future, the AI industry might shift to task-based rather than word-based fees. "Welcome to the Age of AGI."