Elon Musk’s company xAI has officially announced the release of Grok 4 Fast, an updated version of its flagship AI model positioned as a faster and more economical alternative to Grok 4.
Primarily it is designed to provide similar performance with improved efficiency.
The company stated that Grok 4 demonstrates the same level of accuracy while using 40% fewer tokens.
However, these key features mitigate the cost of processing queries by approximately 98% compared to the original model. This makes it accessible to perform a wide variety of tasks.
Grok 4 Fast is specifically built on an innovative approach that combines reasoning and non-reasoning modes.
This approach is similar to strategies used by OpenAI and Anthropic (Claude Opus). It commonly allows the model to automatically switch between in depth analysis and instant response generation.
This Grok 4 Fast ranked first in the LMArena rating for search tasks and eighth in text response quality.
The technical features of the new model are that it supports a context window of up to two million tokens, which remarkably expands its capabilities for analyzing large volumes of information.
The model was significantly trained using action-based learning methods and optimized for fast interaction with external tools such as code execution and web search.
Grok 4 Fast is officially available to end users through the Open Router and Vercel AI Gateway platforms, and on web applications such as iOS and Android.
The initial cost of working with the model starts at $0.20(approximately 17 rubles) per million tokens.