xAI has made its Grok 4.3 language model available on Amazon Bedrock. The model runs on Mantle, Amazon's new inference engine, and is now generally available. With this launch, xAI becomes a model provider on Bedrock for the first time, opening its flagship model to enterprise workflows and AI agents.
The essentials
- Grok 4.3 is now available on Amazon Bedrock using Amazon's new Mantle inference engine
- The model offers a 1 million token context window and supports text and image inputs
- xAI positions Grok as 2–10x more cost-efficient than other frontier models at comparable intelligence levels
- Configurable reasoning effort (none, low, medium, high) per request enables flexible latency-intelligence trade-offs
Why Grok is built for agents and long documents
Grok 4.3 is purpose-built for enterprise workloads where accuracy matters. According to xAI's own benchmarks, the model ranks at the top: it achieved first place on the Artificial Analysis Omniscience benchmark with the lowest hallucination rate among frontier models. It also ranked #1 on the Artificial Analysis Tau2 Telecom benchmark for tool calling in customer support scenarios, as well as on the Vals AI Case Law and Corporate Finance benchmarks for document understanding.
This makes Grok practical for use cases like contract review, credit agreement analysis, and financial document question answering, where the model reasons over long inputs and then calls external systems.
Flexible reasoning levels instead of one-size-fits-all
A key feature: you can control how intensively the model thinks before answering on a per-request basis. Four effort levels (none, low, medium, high) allow a single model to serve the full spectrum of tasks:
| Effort Level | Use Case | Priority |
|---|---|---|
| None | Classifications | Low latency |
| Low/Medium | Standard queries | Balanced |
| High | Contract analysis, case law | Accuracy over speed |
The 1 million token context window accommodates long documents and multi-turn conversations without context loss.
API access via OpenAI-compatible interfaces
Grok runs on Mantle and thus differs in access method from other Bedrock models. Instead of the standard Bedrock Runtime API, it uses OpenAI-compatible APIs. You can invoke Grok 4.3 either with the OpenAI SDK or direct HTTPS requests to the Chat Completions API or Responses API. Region-specific endpoints follow the pattern https://bedrock-mantle.[REGION].api.aws/openai/v1.
This is a pragmatic move: developers already working with OpenAI APIs can integrate Grok with minimal effort.
What this means for enterprises
Grok's availability on Bedrock opens a new option for document processing and agent development. Particularly for financial services, insurance, and legal departments, the combination of long context length, tool calling, and configurable reasoning could prove valuable. However, it remains unclear how Grok 4.3 performs in direct comparison with Claude 3.5 or GPT-4o in real-world scenarios – and whether the promised cost efficiency holds up in practice. Organizations will also need to assess whether the OpenAI-compatible API structure fits their existing AWS workflows.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
