SNACK three-line summary
- xAI said Grok 4.3 became generally available on Amazon Bedrock on June 17. Grok can now be called directly inside AWS, rather than only through xAI’s own route.
- The key specs are a 1 million-token context window and adjustable reasoning effort. xAI says teams can control the amount of reasoning through
none,low,medium, andhighlevels. - Pricing was also disclosed right away. At $1.25 per 1 million input tokens and $2.50 per 1 million output tokens, teams running many long-document or agent workflows need to revisit their cost calculations first.
Snackgirls editor note
AIKO: “Rather than just ‘another Grok model was added,’ the bigger point seems to be that there is now a path to evaluate Grok within AWS procurement, security, and permission systems.”
Red: “It is not only about looking at a model performance table. The practical question is whether it can be plugged into the Bedrock pipeline a team already uses. That is why the change in deployment channel itself is the news here.”
What was announced
xAI announced on June 17 that Grok 4.3 is now generally available on Amazon Bedrock. According to the announcement, developers can now call Grok 4.3 through Bedrock’s inference engine, and AWS also posted a dedicated model card on the same day.
In simpler terms, Grok has moved beyond being “a model you only see on xAI’s site” and has entered AWS as a model that can be considered for enterprise deployment. For teams that care about internal cloud governance and access-control systems, that difference is bigger than it may look.
What can actually be used
The points shared by both xAI’s announcement and the AWS model card are a 1 million-token context window, adjustable reasoning effort, and support for tool calling, structured output, and streaming. AWS documentation lists the model ID as xai.grok-4.3 and explains that it can be accessed programmatically through an endpoint for Bedrock Mantle.
The pricing also arrived in concrete numbers. It is $1.25 per 1 million input tokens and $2.50 per 1 million output tokens. That may look inexpensive at first glance, but if teams frequently run long codebases, long contracts, or multi-turn agent tasks, accumulated token usage can grow quickly.
Why it matters
The core of this change is not a performance boast, but the deployment channel. For organizations already using AWS Bedrock as their standard route, adding a new model may involve less separate vendor onboarding and create room to test Grok inside existing cloud operations workflows.
In other words, the conversation moves beyond “people say Grok is good” and into whether it can actually be paired with procurement, access permissions, audit logs, and security reviews in enterprise environments. That is what makes this a more practical shift than a general web release.
What to watch carefully now
xAI strongly emphasizes hallucination rates and benchmark advantages in its official post, but it is safer to treat those as official vendor claims. Real workplace fit can vary depending on each team’s data, prompts, and tool-integration methods.
AWS documentation also separately explains supported regions, service tiers, and endpoint configuration. So before adoption, the most practical first step is to check whether it is available in your region, whether the token cost is manageable, and whether it conflicts with your existing Bedrock workflow.
Sources and checked date · Announcement 2026-06-17 / checked 2026-06-18T01:24:15+00:00
Sources
Leave a comment