Overview and Model Capabilities
xAI’s Grok 4.7 is available on Amazon Bedrock, adding a frontier model designed for coding, long-running agents, and knowledge work to the managed service catalog. Developed by xAI, the model introduces a 500K token context window alongside four configurable reasoning effort levels. Cloud architects and engineering teams can integrate the model directly into existing AWS architectures to handle tasks that require extended context retention.
According to xAI, Grok 4.7 is its most capable model for coding and knowledge work, with an emphasis on endurance rather than raw speed, working longer on difficult tasks and checking its own work more carefully before moving on. According to xAI, the company trained the model to natively understand the Grok Bot harness, resulting in improvements in conversational tasks and general knowledge work.
Architecture and Cross-Region Inference Profiles
Deployment on Amazon Bedrock relies on cross-Region inference profiles to route traffic efficiently across commercial infrastructure. Developers target the bedrock-runtime endpoint using two primary identifiers. The US geographic profile, designated as us.xai.grok-4.7, keeps processing within the US geography, addressing US data residency requirements. Alternatively, the global profile, designated as global.xai.grok-4.7, routes each request to any supported commercial AWS Region, spreading load across more capacity, and is priced below a geographic profile.
API Integration and Authentication
Integration with Grok 4.7 accommodates multiple developer workflows through broad API support. The model supports the Responses API, the Chat Completions API, InvokeModel, and the Converse API. For teams migrating existing applications built around third-party toolchains, 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 AWS Identity and Access Management credentials. The AWS SDKs reach the same model through Converse, signing requests with ordinary AWS credentials. Additionally, xAI first released Grok 4.7 on September 21, 2026, in a launch announcement that made the model available in Cursor and Grok Build, through the Grok API, via third-party coding harnesses, and on model routers and cloud platforms.
Configurable Reasoning Effort and Token Handling
Reasoning is always active on Grok 4.7. Developers can adjust this behavior using four configurable reasoning effort levels: low, medium, high, and xhigh. Effort is set through the reasoning parameter on the Responses API or through additionalModelRequestFields on Converse, with high as the default.
Multi-turn workflows benefit from encrypted reasoning handling. Reasoning content is encrypted; it can be returned by passing include: “reasoning.encrypted_content” on a Responses API request and then sent back on subsequent turns to give the model its own prior reasoning as context in a multi-turn conversation. The Chat Completions API does not return reasoning tokens.
Security, Guardrails, and Operational Controls
Operational security and cost management features are integrated directly into the Bedrock deployment. Implicit prompt caching applies automatically to repeated prompt prefixes, so agents that resend a large system prompt or reference document on every turn pay the cached rate for that prefix. It also supports Guardrails, prompt caching, and structured outputs on Amazon Bedrock. Bedrock Guardrails attach by ID and version on the request, applying content filters, denied topics, personally identifiable information redaction, and word policies to both the prompt and the response.
Data governance is further supported through Amazon CloudWatch. Structured outputs constrain a response to a JSON Schema so downstream code can parse it directly, and invocation logging captures each call in Amazon CloudWatch with the request, the response, and token counts including reasoning tokens.
Service Tiers, Pricing, and Training Context
Amazon Bedrock offers three service tiers for Grok 4.7. Standard is pay-per-token with no commitment; Priority delivers faster, prioritized processing for a premium; and Flex provides lower-cost access for work that is not time-sensitive.
According to xAI, Grok 4.7 uses a new, larger base model trained through a longer reinforcement learning run over a harder mix of tasks, weighted toward problems that take many hours to complete. In that announcement, xAI priced the model starting at $2 per million input tokens and $6 per million output tokens, and said it serves a fast variant at twice the output speed for twice the price. Grok 4.7 is also available in Amp.
Evaluations, Benchmarks, and Safety Research
Published evaluations for Grok 4.7 span software engineering with CursorBench and DeepSWE, multi-hour terminal and office work with Terminal-Bench and AA Briefcase, electrical engineering with EEBench, legal work with the Harvey Legal Agent Benchmark, and clinical reasoning with HealthBench Professional. xAI further reports stronger document and presentation generation and describes gains on professional knowledge work of the kind done by lawyers, nurses, and financial analysts. According to xAI, Grok 4.7 was built with an entirely new safeguard stack and is the strongest model it has tested on refusals and jailbreak resistance. On cyber security specifically, xAI reports the model allows only a small fraction of risky dual-use prompts through while rarely blocking legitimate security work, and it has begun giving selected cyber security partners invite-only access to Grok 4.7’s red-team capabilities for defense research.
According to Artificial Analysis evaluation results cited by AWS, Grok 4.7 records an Intelligence Index of 46 against 44 for Grok 4.6, a Coding Agent Index of 56 against 47, an AA-Briefcase Elo of 1,657 against 1,546, a GDPval-AA Elo of 1,695 against 1,605, and an AA-Omniscience Index of 32 against 30, with a hallucination rate of 29% against 34%. Artificial Analysis measured Grok 4.7 at its xhigh reasoning effort and Grok 4.6 at the effort level it reported for each measure. The post flags the tradeoff in the final row of that table: Grok 4.7 used roughly 81,000 output tokens per Intelligence Index task, against about 38,000 for Grok 4.6, which the post characterizes as roughly double per task and a reason to set the effort level deliberately rather than inherit the default.
Sources
- Grok 4.7 is now available on Amazon Bedrock
- AWS Adds xAI’s Grok 4.7 to Amazon Bedrock With 500K Context Window – Unite.AI
- Grok 4.7 is now available on Amazon Bedrock and in Amp
Explore more AI Industry coverage.
