xAI's Grok 4.7 lands on Amazon Bedrock with 500K context and four reasoning tiers

Grok 4.7, which hit xAI's public API on September 21, is now generally available on Amazon Bedrock. Access runs through US Geo and Global cross-Region inference profiles. Key specs:

- a 500,000-token context window;

- four configurable reasoning effort tiers, from low to xhigh;

- compatibility with the Responses, Chat Completions and Bedrock Converse APIs, so teams can drop it into existing Bedrock agent stacks.

On capability, AWS says Grok 4.7 improves over Grok 4.6 on mixed-document handling and on repo-scale coding with planning and error recovery. It also cites gains for browser-use agents. According to xAI, the model sits on a new 2.1-trillion-parameter base, about 40% larger than Grok 4.6's 1.5T. It was trained with a longer reinforcement-learning run on a harder task mix and scored 46.3% on CursorBench 4.0. xAI's own API prices it at $2 input and $6 output per million tokens.

For AWS, the launch fits its week-long 'model choice' narrative. The weekly roundup also added GPT-6 Sol, GPT-6 Luna and Claude Opus 5.5 to Bedrock, and it emphasized picking the right model per step by cost and latency. For xAI, Bedrock brings enterprise procurement channels it lacks on its own, as it pushes new Team Bots for workplace deployment.

Developers have raised caveats. They criticize the '200k cliff': crossing 200K tokens doubles the rate on the entire request rather than only the excess. Teams upgrading from Grok 4.5 or 4.6 warn that the pricing, context and reasoning-control changes can break existing integrations. A rebranded 'SpaceXAI Docs' header also raised eyebrows. Watch whether Bedrock pricing mirrors xAI's direct rates and how Grok 4.7 fares against Sonnet 5.5 in independent agentic benchmarks.