Grok 4.7 is now available on Amazon Bedrock
AWS announced xAI's Grok 4.7 on Bedrock on September 28, 2026, with a 500K token context window and four configurable reasoning effort levels.
AWS announced that xAI's Grok 4.7 is now available on Amazon Bedrock. The post is dated 28 SEP 2026. AWS describes it as a frontier model built for coding, long-running agents, and knowledge work, added to the Bedrock model catalog. AWS says xAI introduced Grok 4.7 in a launch announcement published September 21, 2026.
What is available
According to the AWS post, Grok 4.7 offers a 500K token context window and supports configurable reasoning effort at four levels: low, medium, high and xhigh. It accepts text and image input and returns text.
Grok 4.7 is served on the bedrock-runtime endpoint through cross-Region inference profiles, and it supports the Responses, Chat Completions and Converse APIs. Two profiles are listed:
us.xai.grok-4.7, a geographic cross-Region profile. AWS says it keeps processing within the US geography, which addresses US data residency requirements.global.xai.grok-4.7, a global cross-Region profile. AWS says it routes each request to any supported commercial AWS Region and is priced below a geographic profile. The tradeoff, AWS writes, is less control over where a given request is served, which can mean more variable latency.
How to access it
The OpenAI SDK works against the /openai/v1 path with a bearer token, which can be either an Amazon Bedrock API key or a short-term token minted from your AWS Identity and Access Management (IAM) credentials. AWS SDKs reach the model through Converse with ordinary AWS credentials. Bearer-token authentication additionally requires the bedrock:CallWithBearerToken permission.
Before the first call, AWS tells users to confirm in the Bedrock console that the model is available for the AWS Region they plan to use.
Service tiers and platform features
The post lists three service tiers: Standard, which is pay-per-token with no commitment; Priority, which delivers faster, prioritized processing for a premium; and Flex, which costs less for work that is not time-sensitive. Bedrock features listed for the model include implicit prompt caching, Guardrails, structured outputs with JSON Schema, and invocation logging in CloudWatch.
Claims from xAI and Artificial Analysis
These figures are relayed by AWS and are not independently verified. According to xAI, Grok 4.7 uses a new and larger base model. Per Artificial Analysis, as cited by AWS, Grok 4.7 at xhigh effort scores 46 on the Intelligence Index against 44 for Grok 4.6, and 56 against 47 on the Coding Agent Index. The same table puts its AA-Omniscience hallucination rate at 29% against 34%.
AWS itself notes the cost of those gains: they come with roughly double the output tokens per task (about 81k against about 38k per Intelligence Index task), so the effort level should be set deliberately.
What we don't know yet
- Which specific AWS Regions can serve the model. The post says "any supported commercial AWS Region" for the global profile and "US geography" for the US profile, without listing Regions.
- Prices. The post does not state them and points to the Bedrock pricing page; the size of the Priority premium and the Flex discount is also not stated.
- Maximum output tokens, rate limits and quotas.
- The reasoning effort level at which Grok 4.6 was measured. AWS says only that it is the level Artificial Analysis reported for each measure, so the comparison may not be like-for-like.
- Independent reproduction of the benchmark figures above.
Source: AWS