Grok 4.6 is here: xAI’s model for coding, agents, and complex work

xAI released Grok 4.6 on August 12 as its new flagship for software engineering, long-running agents, and difficult reasoning. It offers a 500K-token context window, configurable reasoning effort, and tool support.

What changed

The focus is sustained work rather than chat alone: Grok 4.6 can keep a goal across longer tasks, call functions, search current information, and navigate large codebases. xAI recommends the same model for both general chat and coding.

Grok 4.6 API pricing

xAI and OpenRouter list the same base rates: $2 per million input tokens and $6 per million output tokens. KIE lists 160 and 480 credits, equal to $0.80 and $2.40. NeuralSpace therefore routes Grok 4.6 through the cheaper KIE channel while billing users from actual consumption.

Where to try it

Grok 4.6 is available in NeuralSpace Chat, in Code, and through the OpenAI-compatible public API under grok-4-6. Reasoning effort can be set from low to xhigh.

Low or medium is a sensible default for short requests. For architecture, debugging, and multi-step agent workflows, start at medium or high and increase it only when quality matters more than latency.