Amazon Web Services (AWS) announced on September 28, 2026, the availability of xAI's Grok 4.7 model on Amazon Bedrock. This new frontier model is designed for advanced generative AI applications, particularly excelling in coding, long-running agents, and complex knowledge work. Grok 4.7 offers a 500,000 token context window and supports configurable reasoning effort across four levels: low, medium, high, and xhigh.

The introduction of Grok 4.7 on Bedrock provides developers with an upgraded tool for building more robust and safer AI applications. According to xAI, Grok 4.7 is their most capable model for coding and knowledge work, distinguished by its ability to "work longer on difficult tasks and verify its own output more carefully before moving on." This release builds upon the capabilities of previous versions, Grok 4.6 and Grok 4.3, offering enhanced self-verification and improved context management within the managed infrastructure of Amazon Bedrock.

Grok 4.7's 'Endurance' and 'Self-Checking' for Advanced AI Tasks

Grok 4.7 introduces significant advancements in how large language models handle complex and extended tasks, primarily through its enhanced "endurance" and "self-checking" capabilities. Endurance refers to the model's improved ability to manage longer contexts and persist through multi-step problems, while self-checking involves more rigorous internal verification of its own outputs before proceeding.

These improvements stem from Grok 4.7's training on a new, larger base model compared to Grok 4.6, coupled with a longer reinforcement learning run focused on a harder mix of tasks, including those that require many hours to complete. For developers, this translates into practical benefits for demanding applications such as intricate coding projects, where the model can produce stronger first passes and iterate more effectively. Independent evaluations by Artificial Analysis indicate that Grok 4.7 shows improvements across its evaluation suite, with the most substantial gains observed in long-horizon agentic knowledge work and coding agents.

On benchmarks like CursorBench 4.0, which specifically tests longer-running coding tasks, Grok 4.7 demonstrates frontier-level performance in price-performance. Furthermore, in professional knowledge-work benchmarks such as GDPval, Grok 4.7 (at xhigh effort) scored 1,695, surpassing Grok 4.6 (high) at 1,605, highlighting its enhanced reasoning and knowledge capabilities for complex problem-solving.

Beyond performance, Grok 4.7 incorporates an entirely new safeguard stack, which xAI reports makes it their strongest model tested for refusals and resistance to jailbreaks. This focus on safety is particularly relevant for dual-use domains like cybersecurity, where the model aims to remain useful for legitimate tasks while effectively refusing dangerous ones. xAI notes that Grok 4.7 allows only a small fraction of risky dual-use prompts while rarely blocking legitimate security work.

Grok Model Comparison on Amazon Bedrock

Developers evaluating xAI models on Amazon Bedrock now have three distinct versions to consider: Grok 4.7, Grok 4.6, and Grok 4.3. While all offer powerful generative AI capabilities, each has specific strengths and integration nuances.

Comparative Features of Grok Models on Amazon Bedrock
Model Version Key Capabilities Context Window Reasoning Effort Levels Safety Features Primary Bedrock Endpoint Integration Notes
Grok 4.7 xAI's most capable model for coding, long-running agents, and knowledge work, featuring enhanced self-verification and longer context management ('endurance'). 500,000 tokens Low, Medium, High, xHigh New safeguard stack, strongest on refusals and jailbreak resistance. bedrock-runtime Supports Responses, Chat Completions, and Converse APIs. Uses cross-Region inference profiles.
Grok 4.6 Built for long-running agents, coding, and knowledge work, with improved safeguards and support for visual/interactive projects. 500,000 tokens Low, Medium, High, xHigh Improved and calibrated safeguards, backed by extensive pre-deployment and third-party testing. bedrock-mantle, bedrock-runtime Offers choice between OpenAI-compatible Mantle and Runtime endpoints. Supports image input and prompt caching.
Grok 4.3 Reasoning-first model well-suited for enterprise workloads like customer support, web development, and financial Q&A, delivering consistent results across conversational AI. Not specified (earlier version) None, Low, Medium, High Not specified (earlier version) bedrock-mantle Initial model on Bedrock, reachable through Mantle inference engine. Offers strong tool use and instruction-following.

Grok 4.7 and Grok 4.6 share the same 500,000 token context window and four configurable reasoning effort levels (low, medium, high, xhigh) on Amazon Bedrock. However, Grok 4.7's primary advancements over Grok 4.6 are in its enhanced self-verification and qualitative improvements in how it utilizes its context for "endurance" on difficult tasks, rather than an increase in raw context window size. Grok 4.3, the earliest version, was initially available on the Bedrock Mantle endpoint, while Grok 4.6 offered both Mantle and Runtime endpoints. Grok 4.7 is primarily served on the `bedrock-runtime` endpoint.

Integrating Grok 4.7 into Amazon Bedrock: Practical Guidance

Integrating Grok 4.7 into Amazon Bedrock involves understanding its specific endpoint, authentication methods, and how to leverage Bedrock-specific features. Grok 4.7 is served on the `bedrock-runtime` endpoint through cross-Region inference profiles, supporting the Responses, Chat Completions, and Converse APIs.

For authentication, developers have two primary paths. The AWS SDKs can be used to reach the model through the Converse API, signing requests with standard AWS Identity and Access Management (IAM) credentials. Alternatively, 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 generated from AWS IAM credentials. Developers porting existing integrations might find the OpenAI SDK more convenient, while those seeking a consistent message shape across models in their account, along with invocation logging and response streaming through standard Bedrock event types, may prefer the Converse API with AWS SDKs.

When configuring IAM policies for Grok 4.7, it is essential to allow `bedrock:InvokeModel` for specific inference profiles (e.g., `us.xai.grok-4.7` or `global.xai.grok-4.7`) and `bedrock:CallWithBearerToken` if using bearer token authentication. It is important to note that profiles are scoped individually, meaning a policy for `us.xai.grok-4.7` will not cover `global.xai.grok-4.7`.

Bedrock offers several features that enhance Grok 4.7 integration, including PrivateLink for secure network connectivity and Guardrails for content moderation. These AWS-native features provide additional security and control that are not inherently available with direct API access. For long, unattended agent runs, AWS specifically highlights the utility of CloudWatch invocation logs and structured JSON Schema outputs. Developers should also consider the configurable reasoning effort levels (low, medium, high, xhigh) and benchmark them against their specific workloads to optimize for cost and performance, reserving higher effort levels for planning steps and long agent trajectories where early mistakes can compound.

Decision: Choosing the Right Grok Model for Your Bedrock Project

Developers should evaluate Grok 4.7 for new projects or upgrades requiring enhanced self-verification and 'endurance' for complex coding and knowledge work, leveraging its bedrock-runtime endpoint integration, while considering Grok 4.6 for existing projects that benefit from its dual endpoint support or Grok 4.3 for established enterprise workloads on Mantle.

A measurable indicator of successful adoption would be observed improvements in task completion rates for long-running coding or knowledge tasks compared to previous Grok versions, or optimized cost-performance for specific reasoning effort levels.

Sources