xAI's Grok 4.3 lands on Amazon Bedrock with 1M token context

xAI’s Grok 4.3 is now generally available on Amazon Bedrock, giving enterprise teams a powerful new option for building agents and complex AI workflows. The model boasts a 1 million-token context window, configurable reasoning effort (none, low, medium, high), and strong performance in tool calling and instruction following — all critical for agentic workloads. According to xAI, Grok 4.3 ranks #1 on the Artificial Analysis Omniscience benchmark with the lowest hallucination rate among compared frontier models. It also leads on the Tau2 Telecom benchmark for tool calling in customer support and on Vals AI benchmarks for legal and financial document understanding. xAI claims the model sits on the intelligence-versus-cost Pareto frontier, delivering 2 to 10 times more intelligence per dollar than other top models.

Accessing Grok 4.3 on Bedrock is done through the Mantle inference engine, which uses OpenAI-compatible APIs. Developers can invoke the model via the Chat Completions API or Responses API, using either short-term bearer tokens (recommended for production) or long-term Amazon Bedrock API keys (for exploration). The default settings differ from OpenAI's standard: temperature defaults to 0.7, top_p to 0.95, and max_completion_tokens to 131,072. With its mix of low-latency options (none effort) and deep reasoning (high effort), Grok 4.3 is positioned as a single model that can handle everything from simple classification to complex contract analysis and multi-turn agent sessions.