Grok 4.3 arrives on Amazon Bedrock for businesses

Grok 4.3 is now available to all developers in supported AWS regions through Amazon Bedrock, Amazon’s platform for using AI models with managed, secure infrastructure.

The announcement comes from xAI, the company behind Grok. Its arrival allows teams and businesses to integrate the model into their applications without having to operate all the infrastructure required to run it themselves.

What Grok 4.3 offers

Grok 4.3 includes a 1 million token context window. Context is the amount of information a model can process at once, so this capability makes it possible to work with very long documents, large codebases, or multiple related files without splitting them into as many parts.

You can also choose how much effort the model puts into reasoning before responding:

• none: prioritizes speed for simple requests.
• low: adds a limited level of reasoning.
• medium: balances speed and analysis.
• high: dedicates more resources to complex problems.

In practice, an internal assistant could quickly answer a question about a document using none, then use high to analyze a long contract, review code, or solve a problem that requires multiple steps.

The figures presented by xAI

According to xAI, Grok 4.3 achieves the lowest hallucination rate among advanced models, meaning it produces fewer responses containing invented or incorrect information. The claim comes from the company itself, and the announcement does not detail the test or dataset used for the comparison.

xAI also says the model sits on what is known as the Pareto frontier between intelligence and cost. In simple terms, it claims to offer a better balance between capability and price than other alternatives. Its estimate is between 2 and 10 times more intelligence per dollar than other advanced models.

That figure does not mean every task is between 2 and 10 times better or cheaper. It is an aggregate comparison of performance and cost, and the result depends on the tests, the type of use, and the configuration selected.

What changes for you

If you are developing an AI application on AWS, you can now try Grok 4.3 from the Amazon Bedrock environment and adjust usage based on the difficulty of each request. That can help control costs and response times: there is no reason to use the highest reasoning level for every question.

The ability to process up to 1 million tokens can also reduce the need to summarize or split information before sending it to the model. Still, more context does not guarantee correct answers: the data must be well organized, and important responses need to be reviewed.

Grok 4.3’s general availability puts the focus on two practical decisions for businesses: how much context their applications need and how much reasoning they are willing to pay for on each task. The next point worth watching is whether independent tests confirm the accuracy and cost figures presented by xAI.