SpaceXAI releases Grok 4.6 for long-running agents

Background

Grok 4.5 had been released a month earlier, on 16 July 2026, as xAI's coding and agent-focused model. Grok 4.6 came as the next step in the same line.

What happened

On 12 August 2026 the company released Grok 4.6^[1]. It focuses on long-running agents and multi-step work such as research, codebase tasks and building applications^[1].

Key details

The company said Grok 4.6 scored 61 on the Artificial Analysis Intelligence Index, up from 56 for Grok 4.5^[1]. The model has a 500,000-token context window and introduced an xhigh reasoning-effort level^[2].

Pricing stayed at $2 per million input tokens and $6 per million output tokens^[1]. The model launched in Cursor, Grok Build, the xAI API and partner platforms^[1]. It was also offered through GitHub Copilot and Amazon Bedrock^[1][3].

Impact

The release extended the Grok line into long-running agent work without changing the base prices from the previous model^[1]. It was available across the main coding tools and cloud channels that xAI listed at launch^[1][3].

For readers, a context window is the total amount of text, including the prompt and the reply, that a model can handle at one time. The 500,000-token window is large enough for a very long document or a large codebase in a single request, which suits the release's focus on codebase tasks^[1][2]. Prices are charged per million tokens, so a 100,000-token prompt at $2 per million input tokens costs about $0.20. Higher reasoning effort generally means the model spends more time working before it answers, which can help on hard tasks but makes responses slower.

The details in this entry come from xAI's announcement and its release notes^[1][2]. The main points to remember are the release date, the stated focus on long-running agents, the unchanged prices, and the list of platforms where the model could be used^[1][3].