Introducing Grok on Amazon Bedrock

The launch of Grok 4.3 on Amazon Bedrock marks a significant step in providing enterprise-grade AI models optimized for reasoning and long-input processing. Developed by xAI, Grok 4.3 is designed to handle complex tasks such as document understanding, contract review, and customer support, where accuracy and reasoning depth are critical. It features a 1 million token context window, enabling it to process lengthy documents and multi-turn conversations effectively.

Grok 4.3 is built on Mantle, Amazon Bedrock's next-generation inference engine, and is compatible with OpenAI APIs, allowing developers to integrate it seamlessly into existing workflows. It offers configurable reasoning effort levels, from none to high, to balance latency and depth of reasoning. The model excels in tool calling, which is essential for agentic applications that require external system interactions.

According to xAI, Grok 4.3 outperforms industry benchmarks, ranking first on several tests, including the Artificial Analysis Omniscience benchmark and the Vals AI Case Law benchmark. It also demonstrates a high level of token efficiency, providing more intelligence per dollar compared to other models. Access to Grok 4.3 can be achieved via the OpenAI SDK or direct API calls, with options for secure authentication using AWS credentials.

This release is expected to impact AI development by enabling more reliable and scalable enterprise applications, especially in areas requiring detailed reasoning and long document analysis. It also enhances the capabilities of Amazon Bedrock as a platform for deploying advanced foundation models, supporting a broader range of AI use cases.