Grok 4.6 is xAI's newest flagship model, released in August 2026. It replaces Grok 4.5 as the recommended default. Here is what actually changed and what it costs.
1. What is Grok 4.6?
Grok 4.6 is the newest model from xAI, released in August 2026. The official xAI developer docs now list it and recommend it as the default, describing it as the most intelligent and fastest model xAI has built for chat and code.
In practice it replaces Grok 4.5 as the flagship. Grok 4.5 remains available, and the older Grok 4.3 is still listed for people who want a lower price or a larger context window.
2. What changed from Grok 4.5?
The headline change is capability, not price. On the official pricing page, Grok 4.6 carries the same rates as Grok 4.5 and the same 500k token context window. So this is positioned as a quality upgrade at no extra cost rather than a new pricing tier.
Independent coverage described Grok 4.6 as a post-training upgrade that raised its scores on public leaderboards. Treat specific benchmark figures as reported claims and check the source before relying on them, because leaderboard positions move quickly as new models launch.
Grok 4.6 costs the same as Grok 4.5 and keeps the same context window, so the reason to move is capability, not price.
3. How much does Grok 4.6 cost?
Through the xAI API, Grok 4.6 is priced at 2 US dollars per 1 million input tokens and 6 US dollars per 1 million output tokens for prompts under 200k tokens. Prompts at or above 200k tokens are billed at the higher long-context rate of 4 dollars input and 12 dollars output per 1 million tokens.
For consumers, Grok is reached through the Grok apps and SuperGrok plans rather than by paying per token. If you use Grok inside the app, the model you get depends on your plan, so check what your plan includes at the official pricing page.
4. Grok 4.6 vs other AI models
The wider pattern in 2026 is that the top models are clustered close together and price has become the main difference. New releases from other labs, including Kimi and the latest GPT and Gemini models, keep trading places at the top of public leaderboards.
Rather than chase a single winner, pick by the job. Grok's strengths are live web and X search and bundled image and video creation. If your work depends on today's context, that mix matters more than a small leaderboard gap. For a neutral side by side, see the comparison pages linked below.
5. Should you switch to Grok 4.6?
If you use the API, switching the model name to Grok 4.6 costs nothing extra and gives you the newer model, so there is little reason to stay on 4.5 unless you have tested and prefer it. Run your own real task through both before you commit a workload.
If you are cost sensitive or need the largest context window, Grok 4.3 is still listed at a lower price with a 1 million token window. The newest model is not always the right one for high-volume or very long-context work.
6. How to use Grok 4.6
Developers select Grok 4.6 by using its model identifier in an API call, the same way earlier Grok models are called. The xAI API docs list the exact identifier and the current rates.
For everyday use, open Grok on the web or in the app while signed in. Model availability in the app follows your plan, so confirm what your plan includes at the official checkout before you rely on a specific model.