Grok 4.6 launches for long-running coding and research agents

What changed in Grok 4.6?

SpaceXAI released Grok 4.6 with a focus on agents that work across longer coding, research and application-building tasks. The model is available through Cursor, Grok Build, SpaceXAI's API and several infrastructure partners. API pricing starts at $2 per million input tokens and $6 per million output tokens, but higher rates apply to prompts at or above 200,000 tokens.

What launched

SpaceXAI released Grok 4.6 on August 12. The company describes it as an upgrade aimed at longer sequences of work: researching unfamiliar topics, analyzing information, navigating codebases and turning product ideas into working applications. SpaceXAI says its own tests showed more self-checking on longer runs and stronger first attempts at interactive and visual projects. Those are vendor observations, not independent guarantees.

Where Grok 4.6 is available

SpaceXAI says Grok 4.6 is available in Cursor and Grok Build, through the SpaceXAI API, and from partners including OpenRouter, Vercel and Cloudflare. Cursor and Grok Build receive two-times included usage for the first week. That temporary allowance should not be read as unlimited or permanently free access.

The pricing needs one important qualifier

Headline API pricing starts at $2 per million input tokens and $6 per million output tokens. SpaceXAI also offers a fast variant at twice the price. VentureBeat's readback of the API documentation adds that prompts at or above 200,000 tokens move to higher rates, so developers should check the live pricing documentation before estimating a long-context workload.

What changed for agent workflows

The practical shift is the model's emphasis on staying with a task across more steps. SpaceXAI says training covered knowledge work, general coding, web development, computer-aided design and other agent environments. That makes Grok 4.6 relevant to teams comparing models inside existing coding and agent platforms, but The AI Guy has not independently tested its reliability, token use or task-completion quality.

How to evaluate the release

Start with the actual workflow: the agent harness, tools, prompt length, caching, retries and review requirements can matter as much as the model's per-token rate. Compare completed-task quality and total run cost on a reversible test before moving a production workflow. Vendor benchmark gains over Grok 4.5 are useful context, not proof that Grok 4.6 leads every competing model.

The bottom line

Grok 4.6 is a meaningful model update for developers already using Cursor, Grok Build or model APIs. Its launch combines wider distribution, a long-running-agent focus and aggressive headline pricing, with two important cautions: long-context requests can cost more, and the vendor's workflow claims still need independent testing.

Related tools and pages

Grok researched profileOpen related page AI coding agentsOpen related page Cursor researched profileOpen related page How The AI Guy evaluates toolsOpen related page

Official sources

SpaceXAI launch announcement ↗Checked 2026-08-13
VentureBeat independent report ↗Checked 2026-08-13

This is researched analysis based on public product information. The AI Guy has not independently benchmarked the feature described here.

FAQ

When was Grok 4.6 released?

SpaceXAI announced Grok 4.6 on August 12, 2026.

Where can developers use Grok 4.6?

SpaceXAI lists Cursor, Grok Build, its API, OpenRouter, Vercel and Cloudflare.

How much does the Grok 4.6 API cost?

Pricing starts at $2 per million input tokens and $6 per million output tokens. The fast variant is twice the price, and higher rates apply when prompts reach 200,000 tokens.