OpenAI Launches GPT‑6 Astra Claiming Step Toward Artificial General Intelligence

OpenAI announced the public launch of GPT‑6 Astra, calling it the first model that “enters the artificial general intelligence era” and noting it follows incremental upgrades from GPT‑5.5 to GPT‑5.6; the company made the model available through its API and cloud partners.

GPT‑6 Astra is now offered on Amazon Bedrock, where developers can access it with pay‑as‑you‑go pricing, integrate it for text generation, code assistance, and complex reasoning, and test it in sandbox environments as described in Amazon’s rollout notice.

Safety officials cite rising scrutiny, with Reuters reporting regulators questioning the model’s refusal mechanisms after OpenAI released a safety overview that outlines alignment testing, mitigation strategies, and reduced refusal rates observed in the GPT‑5.6‑Cyber variant.

GPT‑5.6‑Cyber adds live voice AI and achieves a 95 percent completion rate on advanced cybersecurity tasks while showing fewer safety blocks, according to OpenAI’s performance benchmarks and CNBC’s report on the model’s reduced refusal behavior.