OpenAI unveils gpt-6 astra: a powerful AI model for professional work

OpenAI has launched GPT-6 Astra, a new AI model designed to enhance professional work across various domains. Astra is capable of performing complex tasks such as financial modeling, software engineering, and cybersecurity with high efficiency. It can complete Financial Modeling World Cup challenges four times faster than the winning human competitor, significantly reducing the time analysts spend on model building. The model is already being used by customers to optimize GPUs, detect financial discrepancies, and create branded presentations. Astra was internally tested at OpenAI, where it helped reduce Codex session latency by 25 times. It also excels at adhering to company standards, producing slide decks that align with business formatting requirements. Astra is part of OpenAI's effort to deliver cost-effective AI solutions, with pricing starting at $10 per million input tokens and $50 per million output tokens. It outperforms previous models in Terminal Bench 4.0 tests, achieving 57.9% compared to 37.3% for GPT-5.6 Sol 2. Additionally, Astra has enhanced safety features, reducing unintended outcomes by 89% compared to GPT-5.6 Sol.