Amazon Bedrock has made OpenAI’s GPT-6.1 Sol generally available for coding agents, computer use and professional document workflows. Customers can invoke it through supported Bedrock APIs and control model access with AWS identity policies.
OpenAI says the model matched GPT-6 Astra on the DeepSWE v1.1 software-engineering benchmark at about one-fifth the cost per task. It also beat the strongest reported GPT-6 Sol score by 6.4 percentage points while using a lower reasoning setting. Those are vendor benchmark claims on defined tests, not a guarantee of the same savings for every repository or agent.
The model can be configured for Codex work across repositories, local files, terminals and tests. AWS also presents it for document analysis and multistep business tools, where the total cost depends on both token prices and whether better decisions reduce retries and extra tool calls.
Bedrock supports IAM access controls, CloudTrail auditing and private network endpoints. AWS says inference uses hardware-isolated infrastructure and customer prompts and completions are not used for training. Traffic flagged by automated abuse classifiers may be retained for up to 30 days, however. Customers needing zero data retention must request it through their AWS account team and should confirm supported regions, endpoints and pricing before deployment.