Grok 4.7 now available on Amazon Bedrock

What's new

xAI launched Grok 4.7 on September 21, 2026. On September 28, Amazon Web Services (AWS) announced that the model is now also available on Amazon Bedrock, Amazon's service that lets companies use a range of AI models within the same system. The model can also be used through xAI's own API, Grok Build, Cursor and other coding tools.

Grok 4.7 accepts text and images as input but responds only with text. It has a context window of 500,000 tokens. A token is a small piece of text, and the context window is the amount of text the model can keep track of in a single request. According to Let's Data Science, it works as a per-request working memory. It is not remembered across calls and does not replace a proper archive or a search function.

What's smart about it

According to xAI, as relayed by AWS, the model checks its own output before moving on to the next step. That is especially useful when an AI has to work independently for an extended period. AWS describes the model as one for coding, long-running agents (AI that carries out tasks in multiple steps on its own) and knowledge work.

You can choose how much the model should think things through. There are four levels: low, medium, high (the default) and xhigh. AWS writes that the choice has a major impact on costs. The model can also call external tools, deliver responses in a fixed format (JSON Schema) and reuse repeated chunks of text in the prompt to speed things up.

On Bedrock, companies can choose between three service tiers: Default (pay per token), Priority (faster and more expensive) and Flex (cheaper for tasks that are not urgent). Bedrock Guardrails and logging of token usage in CloudWatch are supported.

Cheaper or better?

The Bedrock price is not stated in AWS' text, which refers readers to the Bedrock pricing page. xAI's own API price is $2 per million input tokens and $6 per million output tokens for prompts under 200,000 tokens. Once a prompt reaches 200,000 tokens, $4 for input and $12 for output per million applies to all tokens in the request. No monthly price for regular users has been provided.

AWS notes that xAI has evaluated the model on CursorBench, DeepSWE, Terminal-Bench, legal work and clinical reasoning, among others. The results are the maker's own. xAI reports improvements over Grok 4.6, but on most benchmarks Grok 4.7 at xhigh is compared with Grok 4.6 at high, so the comparison is not entirely apples to apples. Let's Data Science recommends testing the model on your own tasks. No comparison with the closest competitors has been provided.

What it's good for

The model is aimed at software development, AI agents that work for long stretches, and knowledge work such as analyzing large documents. The large memory makes it possible to feed it long texts or large codebases in one go.

For companies that already use AWS, it is easy to try out. It can be called through several programming interfaces, including one that is OpenAI-compatible, and access is via Bedrock API keys or AWS credentials. For production, AWS recommends short-lived tokens.

On Bedrock, the model runs through the profiles global.xai.grok-4.7 and us.xai.grok-4.7. Whether it can be used in Denmark and the EU is not stated, and AWS refers readers to the Bedrock console for the current list of regions.