xAI's Grok 4.3 arrives on Amazon Bedrock with 1M-token context and configurable reasoning

xAI's Grok 4.3 became available on Amazon Bedrock on June 15, 2026, making the reasoning-first model accessible to enterprise developers through AWS's managed AI infrastructure. The launch gives Grok broad enterprise distribution through Bedrock's multimodel marketplace alongside offerings from Anthropic, Google, and others.

What's new

AWS announced the availability of Grok 4.3 with a 1-million-token context window and configurable reasoning effort — a four-level dial (none, low, medium, high) that lets developers trade latency for depth based on the task at hand. The model runs on Mantle, a new Amazon Bedrock inference engine designed for price-performance, with support for tool calling, structured output, and response streaming.

Pricing on Bedrock's on-demand tier:

• $1.25 per million input tokens
• $2.50 per million output tokens
• $0.20 per million cached input tokens

AWS describes Grok 4.3 as well suited to enterprise workloads including customer support, web development, case law research, and financial document Q&A.

Context

Grok 4.3 is xAI's current production reasoning model. Until now, access required going through the xAI API or grok.com directly. The Bedrock distribution means AWS customers can access Grok 4.3 through existing AWS accounts, IAM policies, and billing without establishing a separate xAI API relationship.

The Bedrock model marketplace expanded significantly in early 2026. OpenAI's GPT-5.4 and GPT-5.5 arrived on Bedrock through an OpenAI-compatible Responses API, and Anthropic's Claude models have long been Bedrock's flagship offering. Grok 4.3's addition fills out the frontier tier: all major frontier providers are now represented on the platform.

Why it matters

Bedrock distribution is strategically significant for xAI. Enterprise buyers tend to consolidate AI spend within existing cloud relationships rather than sign new vendor agreements. Landing on Bedrock makes Grok 4.3 accessible to AWS's large enterprise customer base with no new procurement overhead.

The Mantle inference engine is more than a routing layer. AWS built a dedicated inference backend for Grok 4.3 — a sign of a deeper technical collaboration than a standard API pass-through. Mantle handles tool calling, structured output, and streaming natively, which matters for agentic workloads that rely on reliable function calling and partial responses.

Configurable reasoning effort is Grok 4.3's distinguishing feature on the platform. Most frontier models expose a binary extended-thinking toggle. Grok 4.3's four levels (none, low, medium, high) allow more granular cost control — developers can dial reasoning all the way down for high-volume, low-complexity tasks and up for work that benefits from extended chain-of-thought.

With GPT-5.x, Claude 4.x, Grok 4.3, and Google Gemini all accessible through managed AWS infrastructure, enterprise AI deployments are converging on cloud-native multi-model patterns rather than direct provider relationships. That trend benefits AWS as the common infrastructure layer.