OpenAI's GPT-6 Astra reaches general availability on Amazon Bedrock
AWS made OpenAI's GPT-6 Astra generally available on Amazon Bedrock, giving enterprises access to OpenAI's most capable model within their own governed cloud environment. AWS CEO Matt Garman highlighted on LinkedIn (2,007 likes) that what stands out is 'the complexity of the work this model can take on' — reasoning across competing inputs, working through large document collections, investigating problems across complex codebases, and carrying work across applications and software interfaces. Customers can call Astra directly through Bedrock or configure ChatGPT Work and Codex to use Astra on Bedrock.
Technically, Astra ships with a 1-million-token context window, deeper reasoning, professional writing, and advanced computer/browser-use capabilities, all wrapped in the IAM, VPC, and CloudTrail controls enterprises already run. AWS Chief AI & Technology Officer Matt Wood framed it in a 'field note': 'OpenAI's most capable model now runs where your data lives.' Pricing lands at a premium $10 per million input tokens and $50 per million output — 2.5x the prior Sol generation's $4/$20.
Competitively, GPT-6 Astra's simultaneous availability across Bedrock, Microsoft Copilot, Foundry, and GitHub Copilot cements it as the enterprise default even as AWS restructures Bedrock internally via 'Project Mantle' after reported performance issues. This is a multi-cloud land grab: the same model now runs on both AWS and Azure, turning model choice into a commodity and shifting competition to serving efficiency, tooling, and price.
Developers pushed back on cost. LlamaIndex CEO Jerry Liu warned on LinkedIn that using Astra to parse documents 'will break the bank' at 10c+ per page, prompting his native LlamaParse connector. On Hacker News, developers called Astra 'expensive AF,' noting Fast mode drains allowances 6x faster and questioning whether marginal capability gains justify premium pricing for routine tasks. Watch whether enterprise adoption sticks at this price point or migrates to cheaper open-weight alternatives like DeepSeek.
