OpenAI’s GPT-6 Astra reaches general availability on Amazon Bedrock

OpenAI's GPT-6 Astra, the company's most capable model to date, reached general availability on Amazon Amazon.com, Inc. ↘ 1.34% Bedrock on Tuesday, completing the broadest phase yet of a rollout that began on September 3 with a narrow set of enterprise partners. AWS said in a blog post that customers can now call the model directly through Bedrock APIs or configure ChatGPT Work and Codex to use it.unite+1

The expansion to Bedrock follows five days of staged access during which OpenAI first opened the model to participants in its Daybreak cybersecurity-defender program, then widened availability to ChatGPT Plus, Pro, Business, and Enterprise subscribers, as well as through the OpenAI API and Microsoft ↘ 0.72% Azure.openai+2

Pricing, Access, and Enterprise Controls

OpenAI has set standard API pricing at $10 per million input tokens and $50 per million output tokens, with cached input at $1 per million tokens and a Fast mode that delivers roughly twice the processing speed at twice the standard price. Requests exceeding approximately 272,000 input tokens are billed at higher rates.shattered+1

ChatGPT subscribers do not pay per token; Astra usage draws from existing plan allowances, with additional credits available for purchase. Enterprise administrators must manually enable Astra in workspace settings before their users can access it, a restriction OpenAI tied to the model's elevated safety classification.yottalabs+1

The 'Critical' Cybersecurity Label

What sets this launch apart from a routine model update is the safety classification attached to it. Astra is the first OpenAI model to reach the "Critical" level for cybersecurity capability under the company's Preparedness Framework, meaning it can, when equipped with the right tools, discover previously unknown vulnerabilities and develop exploits across protected systems without continuous human direction.dallasexpress+1

OpenAI said it strengthened safeguards in response, including stricter isolation, encrypted checkpoints, expanded monitoring, and more conservative restrictions for users flagged as high-risk. CNBC reported that the model triggered OpenAI's most advanced internal safety protections before release. On Bedrock specifically, AWS said zero-operator access is enforced at the chip level, meaning its own operators cannot read prompts or completions during inference, and customer data is not used to train models.cnbc+2

Competitive Context

Astra's $10/$50 pricing places it well above recent rivals. xAI's Grok 4.5 launched earlier this year at $2 per million input tokens and $6 per million output tokens, making Astra roughly five to eight times more expensive per token. OpenAI is betting the gap is justified by Astra's reasoning depth and cybersecurity capabilities, including a reported perfect score on ExploitBench and 97.6% on FrontierMath Tier 4.techpillow+2

OpenAI disclosed that Astra can be harder to monitor than its predecessor, GPT-5.6 Sol, noting the model's tighter control over written reasoning on simpler tasks. "Improving monitorability remains a research priority," the company said in its safety overview.dallasexpress+1