Troubleshooting OpenAI model access on Amazon Bedrock: model not visible, or HTTP 401 "not available for this account"

Short description

OpenAI GPT-5.x models are served on the Amazon Bedrock bedrock-mantle endpoint through the OpenAI-compatible Responses API, and they do not appear in the classic Amazon Bedrock model catalog. Most access problems have one of two causes. Either you are looking in the wrong place or Region, or your account is not yet enabled for the model. Use the section that matches your symptom.

Resolution

You cannot see, select, or access the model

OpenAI models are absent from the classic Amazon Bedrock model catalog because they are served on the bedrock-mantle endpoint. Work through the following checks.

1. Open the bedrock-mantle console experience, which is organized into projects, and confirm the model is listed in the model catalog. A checkmark in the console indicates the input and output types that a model supports, not whether the model is enabled for your account.
2. Confirm the model is offered in the Region you are using. Availability differs by model and changes over time, and the same model can be present in one Region and absent in another. For current Regions, see the model card and the regional availability page in the Related information section.
3. Confirm your AWS Identity and Access Management (IAM) principal has the bedrock-mantle permissions. Listing and viewing models rely on the bedrock-mantle list and get actions, and inference relies on bedrock-mantle:CreateInference. Attach a policy that grants these actions, such as the AmazonBedrockMantleInferenceAccess managed policy. A call authenticated with an Amazon Bedrock API key additionally requires bedrock-mantle:CallWithBearerToken.
4. Address the correct base path for the model family. GPT-5.x models are served at https://bedrock-mantle.REGION.api.aws/openai/v1, and the open-weight gpt-oss models are served at https://bedrock-mantle.REGION.api.aws/v1. Replace REGION with a Region where the model is offered. A request sent to a base path that the model does not serve returns HTTP 400 with a validation_error, and a request naming a model that does not exist or a Region where it is not offered returns HTTP 404 with a not_found_error.

To list what your account can call, request the models list at the /v1/models path.

The model is not available for this account (HTTP 401)

A response such as openai.gpt-5.5 is not available for this account means the model is not yet enabled for your account. Open-weight gpt-oss models can work in the same account while a GPT-5.x model returns this error.

This is not a bearer token problem, and it is not resolved by adding AWS Marketplace permissions yourself. The Marketplace actions these models need are already granted by the AmazonBedrockMantleInferenceAccess managed policy and are called by the service on your behalf, which you can confirm in the policy document linked in the Related information section.

Use the following steps.

1. Confirm you are calling a Region where the model is offered. The same account can return this error in one Region and succeed in another. In AWS GovCloud (US), establish access in the linked commercial Region first, then enable the model in GovCloud.
2. If the Region is correct, the account has not yet been granted access or capacity for that model. This is decided at the account level and is re-evaluated over time. Request access or a limit increase through your AWS account team.
3. Continue normal, compliant usage on the account while the request is reviewed.

The error mentions unsupported countries or regions

If the message states that access is not allowed from unsupported countries, regions, or territories, this is an OpenAI supported-country restriction that is tied to the country and billing details registered on the account. Changing the AWS Region does not resolve it.

The error says the bearer token is invalid

An Invalid bearer token message is a separate problem and is unrelated to model eligibility. It means the Amazon Bedrock API key is missing, malformed, expired, or does not belong to the account being called. Keys are created at the account level in the Amazon Bedrock console. For creating and rotating keys, see the API keys page in the Related information section.

Related information

• OpenAI models on Amazon Bedrock
• Regional availability by model
• Amazon Bedrock API keys
• AmazonBedrockMantleInferenceAccess managed policy
• Inference using the Responses API