The Pulse
AWS Adds xAI’s Grok 4.7 to Amazon Bedrock
AWS made xAI’s Grok 4.7 available on Amazon Bedrock on September 28, 2026. The model offers a 500,000-token context window, configurable reasoning and two cross-Region routing options, with different token prices.
AI.info Team · Written with AI ·
AWS added xAI’s Grok 4.7 to Amazon Bedrock on September 28, giving customers access to the model through the cloud service’s cross-Region inference profiles. The release brings a 500,000-token context window and four configurable reasoning levels to Bedrock, alongside support for text and image input. AWS describes Grok 4.7 as aimed at coding, long-running agents and knowledge work.
A 500,000-token window, with reasoning always on
Users can set reasoning effort to low, medium, high or xhigh. Reasoning is active by default, and AWS lists high as the default setting; choosing a lower level can help limit reasoning-token use on tasks that do not need extended work. The setting is available through the Responses API and through additional request fields when using Bedrock’s Converse API.
The model accepts text and images and returns text. AWS supports access through the Responses, Chat Completions, InvokeModel and Converse APIs, so developers can use an OpenAI-compatible interface or Bedrock’s AWS SDK path. That provides options for integrating Grok 4.7 into applications already built around either interface.
Two routes determine where requests go
Grok 4.7 on Bedrock requires a cross-Region inference profile; AWS does not offer it for in-Region inference through the bedrock-runtime endpoint. The profile named us.xai.grok-4.7 routes within the US geography, while global.xai.grok-4.7 can route across supported commercial AWS Regions. AWS says the global option can spread demand across more capacity, but gives customers less control over where an individual request is processed.
The distinction matters for teams with data-residency requirements. The US profile keeps processing within the US geography, while Global is the lower-priced option for workloads without the same geographic constraint. Customers also need to check model availability for the AWS Region from which they plan to invoke it; the model card lists supported Regions and routing options.
Global inference costs less per token
On the Standard tier, Bedrock’s listed rates per million tokens are $2.20 for input and $6.60 for output with Geo cross-Region inference. Global cross-Region inference costs $2 per million input tokens and $6 per million output tokens. Cached input reads are listed at 55 cents per million tokens for Geo and 50 cents for Global.
AWS also offers Priority and Flex service tiers for the model. Priority costs 1.75 times the Standard rate, while Flex costs half the Standard rate. The choice of profile and tier therefore affects both request routing and the bill, in addition to the amount of reasoning the model uses.
Bedrock adds its own operating controls
The integration includes Bedrock features such as prompt caching, Guardrails, structured outputs and invocation logging. AWS says logging can capture requests, responses and token counts, including reasoning tokens, which can help teams examine usage across longer agent runs. Access requires the appropriate permissions for the inference profile and the underlying model.
Grok 4.7 expands Bedrock’s catalog with xAI’s latest model for coding and extended tasks, but it does not run as a single-Region endpoint on the service. Developers must choose a cross-Region profile, select an API and reasoning level, and weigh the lower Global rates against the geographic control offered by the US profile.