xAI has unveiled Grok 4 Fast, a new reasoning model that is said to deliver top-tier performance without the high costs. Here are five key takeaways about the latest launch:
1. Frontier-Level Performance at a Lower Cost
Grok 4 Fast matches Grok 4’s benchmark performance while using 40% fewer thinking tokens on average. Thanks to this efficiency and a lower price per token, the model achieves a 98% reduction in cost to reach the same benchmark scores. Independent testing from Artificial Analysis confirms it leads the industry in price-to-intelligence ratio.
2. Smarter, Faster Search Capabilities
The model is trained end-to-end with tool-use reinforcement learning, which gives it advanced search skills. Plus, the Grok 4 Fast can browse the web and X in real time, ingest multimedia like images and videos, and synthesize information quickly.
3. Strong Arena Rankings Prove Its Edge
In LMArena’s Search Arena, Grok 4 Fast’s search variant ranked #1 with 1163 Elo, beating competitors by a 17-point margin. In the Text Arena, it placed #8 overall, performing on par with larger models like Grok 4-0709 and outperforming similar-sized competitors by a wide margin.
4. Unified Architecture for Speed and Flexibility
Unlike earlier versions that needed separate models for reasoning and quick responses, Grok 4 Fast handles both with a single architecture. According to the company, it can respond instantly to simple questions or shift into long-form reasoning for complex ones, reducing latency and token use. Besides, developers using the xAI API can fine-tune the balance between speed and depth.
5. Available Now for Everyone
Grok 4 Fast is already live on grok.com, iOS, and Android apps, and is accessible in both Fast and Auto modes. Even free users get access to the latest model.
However, that’s not all. The company has confirmed that they are working on future updates and plans to improve it based on user feedback. The company also indicated that it is working on improving the multimodal and agent-like capabilities.
