xAI launches Grok 4 Fast with 40% boost in token efficiency

Image Source: Unsplash

xAI has launched Grok 4 Fast, a new artificial intelligence model that it says delivers 40 per cent greater token efficiency than earlier Grok systems. 

The company said the improvement allows the model to achieve benchmark results comparable to Grok 4 while consuming fewer "thinking tokens."

According to xAI, the increase in token efficiency, combined with lower per-token pricing, translates into a 98 per cent reduction in cost to reach the same performance levels. 

The company cited an independent review by Artificial Analysis, which found Grok 4 Fast offered a state-of-the-art price-to-intelligence ratio compared with other publicly available models.

The model was built using large-scale reinforcement learning to maximise what xAI calls "intelligence density." 

The firm said Grok 4 Fast outperforms Grok 3 Mini across reasoning benchmarks and, unlike earlier systems, does not require separate models for reasoning and non-reasoning tasks. By combining both functions in a single architecture, xAI said it has been able to cut latency and lower token usage.

The company added that Grok 4 Fast includes a 2 million token context window, which enables it to process longer conversations or documents. It also features upgraded browsing and search capabilities across the web and on X. 

According to xAI, the model can follow links, analyse media such as images and videos, and compile results in real time.

The new system is available immediately on grok.com and through the xAI API. 

The company said all users, including those on the free tier, will now have access to Grok 4 Fast.