AWS adds GPT-6.1 Sol to Amazon Bedrock

AWS has made OpenAI’s GPT-6.1 Sol generally available through Amazon Bedrock, giving AWS customers a managed route to use the model from the Bedrock console or supported Bedrock APIs.

The September 29 rollout adds a distribution channel; it is not the model’s original launch. OpenAI’s API changelog separately records a same-day release through its Responses and Chat Completions APIs. The Bedrock release follows AWS’s addition of GPT-6 Sol and Luna to Bedrock.

AWS and OpenAI position GPT-6.1 Sol for complex coding, computer-use and professional workloads. AWS, relaying a result from OpenAI, said the model matched GPT-6 Astra on the DeepSWE v1.1 coding benchmark at roughly one-fifth the cost per task. AWS also said GPT-6.1 Sol exceeded GPT-6 Sol’s best score by 6.4 percentage points while using a lower reasoning effort. The benchmark and cost claims were not independently reproduced in the opened sources.

AWS said Bedrock customers can govern access with IAM policies, audit invocations through CloudTrail, and keep traffic within network boundaries through VPC endpoints powered by AWS PrivateLink. AWS also said inference data is not used for model training and that using GPT-6.1 Sol does not require opting in to data sharing with OpenAI. The launch post said classifier-flagged GPT-6.1 Sol traffic may be retained by AWS for up to 30 days and processed programmatically, with zero data retention available by request. However, AWS’s separate abuse-detection documentation did not list GPT-6.1 Sol among the models subject to that exception.

OpenAI’s direct model documentation lists a 1,050,000-token context window, a 128,000-token maximum output, text input and output, image input, and tool calling through the Responses API. Those specifications describe OpenAI’s direct service and do not establish Bedrock-specific pricing or deployment settings.

At the research cutoff, AWS had not published a GPT-6.1 Sol entry in its linked Bedrock OpenAI model index. The opened AWS sources did not specify supported Regions, Bedrock model or inference-profile IDs, or Bedrock pricing.