<p>xAI has released Grok 4.3, available now on the xAI API with no waitlist.</p>
<h2>What’s new</h2>
<p>Grok 4.3 builds on the 4.20 architecture with improved instruction-following, better context retention across long agentic sessions, native video understanding, and structured document generation. xAI claims roughly a 20% cost reduction versus 4.20.</p>
<p>The model tops the Artificial Analysis leaderboards for agentic tool calling and instruction following, and ranks first on Vals AI benchmarks for enterprise domains including case law and corporate finance.</p>
<p>Features include:</p>
<ul>
<li>1 million token context window</li>
<li>Always-on reasoning (non-configurable)</li>
<li>No output token limit</li>
<li>Function calling and structured outputs</li>
<li>Knowledge cutoff: December 2025</li>
</ul>
<h2>Pricing and availability</h2>
<table>
<thead>
<tr>
<th></th>
<th>Price</th>
</tr>
</thead>
<tbody>
<tr>
<td>Input</td>
<td>$1.25 / million tokens</td>
</tr>
<tr>
<td>Cached input</td>
<td>$0.20 / million tokens</td>
</tr>
<tr>
<td>Output</td>
<td>$2.50 / million tokens</td>
</tr>
</tbody>
</table>
<p>That’s roughly 5-10x cheaper than comparable reasoning models, according to xAI.</p>
<p>The model is available in the <code>us-east-1</code> and <code>eu-west-1</code> regions with rate limits of 1,800 requests per minute and 10 million tokens per minute. Different limits apply for requests exceeding the 200K context window.</p>
<h2>Integrations</h2>
<p>Grok 4.3 is already available through Vercel’s AI Gateway (set the model to <code>xai/grok-4.3</code> in the AI SDK) and in the Kilo Code IDE, CLI, and cloud agents.</p>


Collections

xAI has launched Grok 4.3 on its API with no waitlist. The model builds on the 4.20 architecture and introduces improved instruction-following, better context retention for long agentic sessions, native video understanding, and structured document generation. It features a 1 million token context window, always-on reasoning, no output token limit, and a December 2025 knowledge cutoff. Pricing starts at $1.25/M input tokens and $2.50/M output tokens — roughly 5–10x cheaper than comparable reasoning models per xAI. The model tops Artificial Analysis leaderboards for agentic tool calling and instruction following, and ranks first on Vals AI enterprise benchmarks. It is available in us-east-1 and eu-west-1 regions and already integrated with Vercel's AI Gateway and Kilo Code.

Grok 4.3 is available via API with 1M token context and built-in reasoning