# Grok 4.7 is available on Amazon Bedrock

- Source: MechNext — https://mechnext.eu/en/article/24-grok-4-7-is-available-on-amazon-bedrock
- Published: 2026-09-29
- Updated: 2026-09-29
- Category: Industry News
- Language: en (Markdown in other languages: ?lang=de, ?lang=sk, ?lang=pl, ?lang=zh, ?lang=ru)

> AWS announced xAI’s Grok 4.7 on Amazon Bedrock, listing a 500K token context window, four reasoning effort settings, two inference profiles and several supported APIs.

An AWS Machine Learning Blog post published on September 28, 2026, announced that xAI’s Grok 4.7 is available on Amazon Bedrock. The post sets out the model’s stated inputs and output, its context window, the interfaces through which it can be called, and the inference profiles named for it. Taken together, these details help a reader distinguish the announcement of availability from the choices involved in making a request. The specifications below are attributed to AWS’s post.

### What the model accepts and returns

On Amazon Bedrock, Grok 4.7 accepts text and image input and returns text, according to AWS. The direction of that exchange matters: images are described as input, while the stated output is text. A reader planning a request that includes an image can therefore identify that input type in the announcement without reading the statement as a claim about image output. AWS also gives Grok 4.7 a 500K token context window. That figure describes the stated context size; by itself, it does not establish the length of a particular response or the result of a particular task.

The post lists four reasoning effort levels: low, medium, high and xhigh. High is the default. These names are settings a caller may need to consider when preparing a request, especially if relying on the default would leave the selected level unclear to someone reading the request later. The announcement supplies the levels and the default, but those details alone do not quantify a difference in performance, time or cost between the settings.

### API routes and inference profiles

AWS says Grok 4.7 is served through the bedrock-runtime endpoint using cross-Region inference profiles. The post lists support for the Responses API, Chat Completions API, InvokeModel and Converse API. Those are the named interfaces in the announcement. A reader comparing integration examples should note which one an example uses, because the post gives separate details about authentication and reasoning content for particular routes. Merely listing the interfaces does not make every response detail interchangeable among them.

The two inference profile names supplied are `global.xai.grok-4.7` and `us.xai.grok-4.7`. For the US geographic profile, AWS says processing stays within the US geography. This is a statement about that profile, so the full profile name matters when considering where processing is meant to occur. The post also describes three service tiers, named Standard, Priority and Flex. It identifies the tiers but the fact of their names does not, on its own, define a particular tier’s behavior for an individual request.

### Authentication and reasoning content

For OpenAI-compatible calls, the AWS post says a Bedrock API key can be used as a bearer token. For boto3 Converse calls, it says authentication uses SigV4 with AWS credentials. These are distinct instructions associated with distinct call paths. Someone preparing a client can use that distinction to match the authentication method to the interface being used, rather than treating the two descriptions as a single universal method.

The post identifies another route-specific difference. The Responses API can return encrypted reasoning content when requested, AWS says, while the Chat Completions API does not return reasoning tokens. Thus, a request for encrypted reasoning content through Responses API and an expectation of reasoning tokens from Chat Completions API are different propositions under the announcement. Both interfaces appear on the supported list, but the stated output detail is specific to each one.

For a reader evaluating an integration, the useful sequence is to identify the planned input, select one of the listed interfaces, identify the appropriate inference profile, and decide whether the reasoning effort setting should be explicit. The [AWS Machine Learning Blog announcement](https://aws.amazon.com/blogs/machine-learning/grok-4-7-is-now-available-on-amazon-bedrock) provides the stated model details for those decisions. It establishes availability on Bedrock and lists these configuration points; it does not, from these facts alone, establish the outcome of any particular use.