Grok 4.6 Explained: Two-Tier Pricing and Specs

In short: Grok 4.6 is an AI model from xAI, released on 7 August 2026 and described in xAI's documentation as a "frontier model for coding, agentic tasks, and knowledge work". It costs $2 per million input tokens and $6 per million output for prompts under 200,000 tokens, and double that once a prompt reaches 200,000. It has a 500K-token context window and reads text and images. xAI's newest model is Grok 4.7.

What is Grok 4.6?

Grok 4.6 is a large language model in xAI's Grok family, available through the xAI API and the Grok apps. It followed Grok 4.5 about three weeks later. The API model name is grok-4.6 (xAI model page).

For how models like this work, see how LLMs work.

How much does Grok 4.6 cost?

Grok 4.6 has two price tiers, set by the size of the prompt. From xAI's model page:

xAI's page says requests whose prompt reaches 200k tokens are billed at the higher rate for all tokens in the request - output included. The output rate therefore depends on how long the prompt was, not on how long the answer is.

What does the 200k-token price step mean in practice?

A conversation that grows past 200,000 tokens of history moves every following request to the higher rate. In a long chat or an agent that carries its history forward, each message re-sends everything before it, so the prompt keeps growing. Once it reaches 200,000 tokens, input and output both cost twice as much per token for every later request in that conversation. The context window is 500,000 tokens, so the higher rate covers every prompt between 200,000 and 500,000 tokens. See tokens and context windows explained.

What are Grok 4.6's specs?

xAI's documentation does not list a maximum output length for Grok 4.6.

How does Grok 4.6 compare with Grok 4.5 and 4.7?

All three have a 500,000-token window and the same input and output prices; they differ on cached input.

xAI's models page now names Grok 4.7 for code and chat, calling it "the most capable model we've built".

How does Grok 4.6 rank?

On LMArena's text leaderboard Grok 4.6 has 1455 Elo (rank 72, high effort). Grok 4.5 has 1465 Elo (rank 53) and Grok 4.7 has 1441 Elo (rank 96, xhigh effort). The full ranked table is on our benchmarks page.

Words to know

• Long-context pricing - a higher rate once a prompt passes a size threshold.
• Prompt - everything sent to the model in one request, including conversation history.
• Cached input - repeated input billed at a lower rate.
• Reasoning effort - how much the model thinks before answering.
• Server-side tools - tools such as web search that run on xAI's side during a request.

Related guides

• Grok 4.5 explained
• Tokens and context windows explained
• How to compare AI models