Grok 4.3 is now on Amazon Bedrock, with 1M context and cheap token pricing
xAI has moved Grok 4.3 into Amazon Bedrock general availability. That matters because a lot of serious AI work is already wired through AWS procurement, IAM, logging, and Bedrock routing — so this is less “another chatbot got an update” and more “Grok is now available where enterprise teams actually deploy models.”
The short version: Grok 4.3 is available through Bedrock, comes with a 1-million-token context window, and supports configurable reasoning effort: none, low, medium, or high. xAI is also pricing it aggressively: $1.25 per million input tokens and $2.50 per million output tokens, according to the company’s launch post.
What changed
xAI says Grok 4.3 is now generally available on Amazon Bedrock in supported AWS Regions. The AWS docs now include a dedicated Grok 4.3 model card, which is the practical bit for builders: this is not just a marketing page, it is hooked into Bedrock’s normal model access and API flow.
The model’s positioning is very clear:
- 1M context window for long documents, codebases, legal/finance packs, and agent memory-heavy workflows.
- Configurable reasoning so teams can trade speed and cost against deeper thinking.
- Enterprise-friendly deployment path via Bedrock instead of a separate xAI integration.
- Low headline pricing versus most frontier model APIs.
xAI also claims strong results on hallucination, telecom tool-calling, legal, and corporate finance benchmarks. Treat those as vendor claims until you test them on your own evals, but the combination of long context, Bedrock availability, and pricing is enough to make this worth paying attention to.
Why builders should care
If your stack already uses Bedrock, this lowers the switching cost for trying Grok in production-like workflows. You can put it next to Claude, Nova, Llama, Mistral, or whatever else you already route through Bedrock, then compare on your own tasks instead of building a bespoke integration just to run a pilot.
The most obvious tests:
- Long-context code review across a whole repo or large PR set.
- Customer-support agents where tool use and low hallucination matter more than benchmark theatre.
- Legal, finance, or compliance document workflows with large context packs.
- Cost-sensitive reasoning jobs where you can use
noneorlowmost of the time and reserve heavier reasoning for hard cases.
For founders, the procurement angle is underrated. A model becoming available inside Bedrock can move it from “interesting, but legal/security will slow us down” to “we can evaluate this next sprint.”
Caveats
The launch post is from xAI, so benchmark claims need independent testing. Region availability also matters; if you run in a region where Grok 4.3 is not exposed yet, the announcement is not immediately useful. And while the token price is attractive, real cost will depend on how much you lean into that 1M context window — long context is powerful, but it is also an easy way to burn money if your retrieval and prompting are lazy.
Still, this is a meaningful builder update. Grok is no longer just competing for attention in its own app/API lane. It is now competing inside the default enterprise model marketplace.
Top comments (0)